people’s preferences, perceptions and frequency of using

62
People’s preferences, perceptions and frequency of using search engines to search for images A study submitted in partial fulfillment of the requirements for the degree of Msc Information Systems Management at THE UNIVERSITY OF SHEFFIELD by Junyu Li September 2013

Upload: others

Post on 18-Dec-2021

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: People’s preferences, perceptions and frequency of using

People’s preferences, perceptions and frequency of using search engines to search for images

A study submitted in partial fulfillment of the requirements for the degree of

Msc Information Systems Management

at

THE UNIVERSITY OF SHEFFIELD

by

Junyu Li

September 2013

Page 2: People’s preferences, perceptions and frequency of using

2

Abstract

With the increase growth of image retrieval, it becomes significant to analysing people’s

perceptions on image retrieval. The frequency, perceptions and preferences of people searching

for images are very important to develop search engines. In this research, 84 participants

completed the questionnaire, and all of them are students or staff in the University of Sheffield.

The survey was designed to find out people’s behaviours on searching for images and which

search engine was the popular engine for image search. In the research, the most popular search

engine was Google and then Bing. On average, people would use search engines to search for

images 5 to 10 times per week, and search for job/academic needs 1 to 5 times per week. Most

people use search engine to search for images to prepare presentation, find out what something

looks like and travel destination. They carry out image search for curiosity, entertainment, gaining

knowledge, solving problems and job/academic need. 90.5% participants thought their last image

search was successful and the search engine was easy to operate. Only 9.5% participants thought

it was unsuccessful. Among unsuccessful participants, most people used Google search by image

function and thought it was hard to operate. In the end, participants thought search engine need

to return more relevant images than irrelevant images, improve the quality or size of the

returned images, provide related information about the image, and provide more search options

or improve the interface.

Page 3: People’s preferences, perceptions and frequency of using

3

Acknowledge Firstly, I would like to express my appreciation for the helps offered by my dissertation

Supervisor Dr. Robert Villa, who always gives me great supports. He always encouraged me

and helped me build confidence and gain courage throughout the whole process of finishing the

dissertation.

Secondly, I would like to thank all of the participants who finished the questionnaire and help

me to finish this research, and I would like to thank my friends, because they give me enough

help to conduct this research and write this dissertation, especially they help me to carry out

pilot test successfully.

Finally, I would like to express my gratitude to my family, who always support me to complete this

research project.

Page 4: People’s preferences, perceptions and frequency of using

4

Contents

Abstract ................................................................................................................................. 2

Acknowledge ......................................................................................................................... 3

1. Introduction ................................................................................................................... 7

1.1 Research aim ........................................................................................................................... 7

1.2 Objectives ................................................................................................................................ 8

Section 1: ................................................................................................................................... 8

Section 2: ................................................................................................................................... 8

Section 3: ................................................................................................................................... 8

1.3 Structure ................................................................................................................................. 8

2. Literature review .......................................................................................................... 10

2.1 Introduction .......................................................................................................................... 10

2.2 Image retrieval development ................................................................................................ 10

2.3 Types of image retrieval and limitation ................................................................................. 11

2.3.1 Text based image retrieval ............................................................................................. 11

2.3.2 Content based image retrieval ....................................................................................... 12

2.3.3 Limitations of image search ........................................................................................... 13

2.4 Image search behaviour ........................................................................................................ 13

2.4.1 Professional users’ search behaviours ........................................................................... 14

2.4.2 Search query affects peoples’ search behaviours .......................................................... 15

2.4.3 Similarity of images affects peoples’ search behaviours ................................................ 16

2.5 Conclusion ............................................................................................................................. 16

3. Methodology ................................................................................................................ 17

3.1 Introduction .......................................................................................................................... 17

3.2 Quantitative methodology .................................................................................................... 17

3.3 Qualitative methodology ...................................................................................................... 17

3.2 Critical incident Technique .................................................................................................... 18

3.3 Survey design ........................................................................................................................ 19

3.3.1 Questionnaire questions Q1-Q10................................................................................... 19

3.3.1 Questionnaire questions Q11-Q19................................................................................. 19

3.4 Implement ............................................................................................................................. 20

3.4.1 Participant ...................................................................................................................... 20

3.4.2 Ethical Aspects ............................................................................................................... 21

3.4.3 Pilot test ......................................................................................................................... 21

3.4.3 Questionnaire granting .................................................................................................. 22

4. Result ........................................................................................................................... 23

4.1 Demographic information Q1-Q4 ......................................................................................... 23

4.2 Perceptions regarding searching for images Q5-Q10 ............................................................ 24

Page 5: People’s preferences, perceptions and frequency of using

5

4.2.1 Q5-Q7 frequency of searching for images ..................................................................... 24

4.2.2 Q8 the reason why carry out image search ................................................................... 25

4.2.3 Q9-Q10 which search engine is popular and why .......................................................... 26

4.3 Most recent experience of searching for images .................................................................. 28

4.3.1 Q11 when participant carried out the last image search. .............................................. 28

4.3.2 Q12-Q13 topics and reasons carrying out last image search ......................................... 28

4.3.3 Q14-Q16 search engine and difficulty & satisfaction of search engines ........................ 30

4.3.3 Q17-Q18 successful and unsuccessful with the last image search ................................ 32

4.3.3 Q19 aspects of search engine to improve ...................................................................... 34

5. Discussion .................................................................................................................... 36

6. Conclusion ................................................................................................................... 39

6.1 Summary ............................................................................................................................... 39

6.2 Limitation .............................................................................................................................. 40

6.3 Future study .......................................................................................................................... 40

Reference ............................................................................................................................ 41

Appendix A Questionnaire ................................................................................................... 45

Part One ...................................................................................................................................... 46

Part Two ...................................................................................................................................... 47

Part Three .................................................................................................................................... 49

Appendix B Ethics Proposal .................................................................................................. 53

Appendix C Ethics Information Consent ................................................................................ 59

Appendix D Approval Letter ................................................................................................. 61

Page 6: People’s preferences, perceptions and frequency of using

6

Table Contents

Table 1 Q1-Q2 gender and age scale of people who participated in the survey .................. 23

Table 2 Q3-Q4 Education background and occupation of participants ................................ 24

Table 3 Q5-Q7 Three different frequencies of searching for images. Frequency 1 was on average image search frequency, 2 was search for images in job/study, and 3 was search for images outside job/study. 1-5 means 1-5 times per week, 5-10 means 5-10 times per week, 10-20 means 10–20 times per week, and >20 means more than 20 times per week. ............................................................................................................... 25

Table 4 Q8 Why carry out image search. The table shows the reasons why carried out image

search, and used Likert Scale to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant, and 5 means very important. In the table, Range means how many numbers participants used to measure the reasons. ............................................ 26

Table 5 Q10 the ranking of reasons for choosing image search. In the table, reasons for

choosing search engines were listed, and used Likert Scale to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant, and 5 means very important. In the table, Range means how many numbers participants used to measure the reasons. ... 27

Table 6 Q11 When was the Last time participants searched for images. ............................. 28

Table 7 Q12 Topics participants chose to search for images in their most recent image search

......................................................................................................................................... 29

Table 8 Q13 Reasons participants chose to search for images in their most recent image search .............................................................................................................................. 29

Table 9 Q12-Q13 correlation of topics and reason participant chose to carry out last image

search. In the table, Frequency was used to analysis the correlation of topics and reasons. ........................................................................................................................... 30

Table 10 Q14-Q16 How participants satisfied with the search engine, how difficulty they felt

to operate the search engines, and which search engine were used to carry out the most recent image search ............................................................................................... 32

Table 11 Q17 the possible reasons for participants having a success experience in image

search. In the table, Likert Scale was used to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant, and 5 means very important. ........................... 33

Table 12 Q18 the possible reasons for participants having a failure experience in image

search. In the table, Likert Scale was used to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant, and 5 means very important. ........................... 34

Table 13 Q19 In which aspects, search engines for image search can be improved (multiple

choices can be made) ...................................................................................................... 35

Page 7: People’s preferences, perceptions and frequency of using

7

1. Introduction

As digital technologies have improved rapidly nowadays, the means of multimedia information

retrieve become various. Image retrieve become more significant for users, according to Yoon’s

research (2011), People search images for multiple utilizations, such as using for computer’s

wallpaper, demonstrated related items on slides, finding some unknown item’s pictures and

photos, etc. Image retrieve was not only based on text annotation, but also based on the content

of images. On one hand, image’s visual features can make itself delivered as either

complementary resources with text documents or independent resources. On the other hand,

the visual features of images generate different users’ needs and search behaviours, so it is

crucial to figure out user’s search behaviours and it can also help to improve the quality of image

retrieval. Image indexing and retrieval systems mostly rely on the computation of similarity

measures between images (Tirilly et al, 2012) and how easy to operate search engine would

conduct different search behaviour (Zhang, 2006). Therefore, those factors would also need to

consider when studying on people’s behaviour on image retrieval.

1.1 Research aim

This study aims to investigate how people use image search engines, people’s behaviours on

using image retrieval systems and why people succeed or fail searching for images. In this

research, several questions will be investigated, and overall aim of this research is to find out the

reason why people use image search engine and how people behave when searching for images.

Firstly, investigate the frequency of people using of image search engines and which search

engine is people’s favourite one and why. Then investigate people’s behaviour through image

retrieval, such as which is the way people used to search image, what their perceptions after

search engine giving the results, are whether they satisfied with those search results, if the

results are not satisfied their need, is the reason would be similarity between the search results

or the operation difficulty of search engines, or other unexpected seasons. The aims to

investigate people’s behaviour through image search would be important for building image

search system. The similarity of images could influence people’s need and the degree of

satisfaction with the results after searching certain types of image. University students may be a

specific group in searching image. In assumption, they would use image retrieval both in personal

way and academic way. This research will focus on the group of students in the University of

Sheffield to study their behaviours on image retrieval and to prove above assumptions. All the

questionnaires and Likert Scales are anonymous, and will be accessed by online channel. The

participants are mainly focus on the students of University of Sheffield. Questionnaires and Likert

Page 8: People’s preferences, perceptions and frequency of using

8

Scales are sent by email system in the University of Sheffield. Overall practicalities of this

research are with less ethical risks because recruiting people online is relatively easy and safe,

and the number of participant is only 100 more or less in this research which is not too large to

investigate.

1.2 Objectives

Section 1: To investigate why people use search engine to search for images and the

frequency of using it.

RQ 1.1 How often do people use search engine to search for images and why do they carry out

image search?

RQ 1.2 Which search engine to search for images do people usually use and which search engine

is most popular with people and why?

Section 2: To investigate people’s behaviours on image retrieval.

RQ 2.1 How easy do people feel about operating search engines to search for images?

RQ 2.2 How happy do people feel about the search results?

Section 3: To figure out why people success or fail in image searching.

RQ 3.1 Do people feel successful when conducting an image search, and why they feel

successful?

RQ 3.2 Do people feel failure when conducting an image search, and why they feel

unsuccessful?

RQ 3.3 Is there any aspects of search engines need to improve?

1.3 Structure

In this dissertation, six chapters would be used to explain this research. Firstly, Chapter one is

introduction, in this section, the background, aims, objectives and relevant concepts of this

research will be expressed. Secondly, literature review will be conveyed in Chapter two, in which

critically reviews will contain several areas, such as development of image search, types of image

search, and people’s behaviours on searching for images. After that, Chapter three will address

the methodology applied in this research, which includes the appropriate methodology to collect

data, the way to design the questionnaire, and the implement of those research questions.

Besides, in Chapter four, results of this research after data collecting will be described, and the

Page 9: People’s preferences, perceptions and frequency of using

9

appropriate tables and charts will be used to display the questionnaire data of the research.

Fifthly, Chapter five is discussion part, in which significant and interesting findings will be

discussed, and the main points about data also need to analysis. Finally, the last Chapter is

conclusion. In this section, summary of this research project, limitation of this research and some

recommendations of the future work will be discussed.

Page 10: People’s preferences, perceptions and frequency of using

10

2. Literature review

2.1 Introduction

Image retrieval is a sort of data search, but specialized for image search. To search for images,

users need query term, such as keyword, image file uploaded, or click on search condition. After

that image retrieval systems would return images matched users’ queries. With the rapid

development of computing technologies, all aspects indicate the growth of image search. As the

increased needs of image retrieves, and the difficulties of images annotations, the queries of

images described by users would be a problem for image search (Chen & Rasmussen, 1999). The

famous semantic gap in image retrieves is the discrepancy between the query a user ideally

would and the one it actually could submit to an information retrieval system. Therefore, to rise

up the precise of image index or to reach an agreement on the content search system returned

and the images users required became difficult. Image annotation and search models building

draw lots of interest of researchers, and recently there are increasingly researchers focus on

image retrieval (Rui et al, 1999).

Although, some possible improvements of image search are still imperfect, researchers in

different fields develop different types of image retrieves. One is content-based image retrieval

and another is text-based image retrieval, researchers use both of them to find ‘similar’ images

to match users’ queries. However, researchers can not only develop the methods of image

search, but also understand users’ underlying questions of images access, for example: how

images are described and what information should be offered to users (Chen & Rasmussen,

1999). Besides, similar images are defined as images of the same object or scene, but viewed

different by users under different conditions (Jegou et al, 2008). If researchers would improve the

image retrieval systems, they may focus on users’ underlying requirements through image search.

Therefore, using creative and useful way may solve those problems, but to understand users’

needs and search behaviours would be important to improve image retrieval systems.

2.2 Image retrieval development

According to Yoon (2011), images can not only be delivered directly from context as

complementary resources of text documents, but also can be searched as primary resources with

distinctive needs of users. As a result of popular use of image search, large search engine

companies like Google and Yahoo! built their own image search library and add this function to

their search engine (Zhang et al, 2006). In commercial way, Google built the largest image library,

Page 11: People’s preferences, perceptions and frequency of using

11

so that users can use Google’s image search engine by typing query term, and Google would

match key words with textual annotation to return the ‘similar’ images (Fergus et al 2005). While

for Yahoo!, the same way was applied to its image search function, and this made Yahoo! become

the strongest competitors to Google in the world (Zhang et al, 2006). Recently, new website like

Flickr, which focus only on social network sites’ photos uploading, develops rapidly because of

the huge amount of image needs required by customers in this area. Although, website like Flickr

own millions of images in its database, but it is uneasy for customer to find out the exact one

which they ‘just know’ existed in those search engine database (Neal, 2009). Customer may be

still unsatisfied with those search engines to meet their distinctive needs. In Rieh and Hilligoss

research (2007), credibility would be concerned mostly in image retrieval, because of its effort

level and strategies which people applied on information seeking. However, for college students,

although credibility was one important factor to considerate during information seeking, they

often compromised information credibility for speed and convenience, especially when the

information retrieval was less important. The same way maybe also apply on the image retrieval

that people would not so critically need the credibility when seeking some less consequential

images. Therefore, this would be the reason why people choose to use Google or Yahoo! instead

of professional image retrieval engine, such as Gettyimages, Tineye or Iamgeseek websites.

2.3 Types of image retrieval and limitation

The technologies of image retrievals have developed rapidly in past 20 years, since 1990’s. This

increased search methods offer new ways to solve the semantic gap of images. Although the

problem is still there, users’ needs are gradually satisfied. There are two major search methods.

One is the traditional concept-based indexing, which searches for images by using both daily and

professional words and terms as query conditions, or describing what an images is or what it is

about in natural language. Another is newly developed techniques called content-based image

retrieval, which is rely on the analysis of image content through pixel-level of images (Chen &

Rasmussen, 1999).

2.3.1 Text based image retrieval

Usually, people use image search engine with composing a text query and executes this query on

intermediary image databases (Villa et al, 2010). The traditional concept-based or text-based

indexing is widely used in commercial search engines. The method called textual annotations for

image retrieval means adding a note or text manually to an image so that system can match

user’s query with those note. It may employ keywords, subject headings, captions, or natural

language text to annotate images (Chen and Rasmussen, 1999). According to Villa’s research

Page 12: People’s preferences, perceptions and frequency of using

12

(2011), this is the standard method of online image seeking, and it is applied widely for online

image involved websites such as Flickr. This type of image search required a large set of human

labelled annotation. In 2001, Jorgensen evaluating a conceptual structure for image content

depiction. They categorise visual content into a “pyramid” with 4 syntactic levels

(type/technique, global distribution, local structure, composition) and 6 semantic levels (generic,

specific, and abstract levels of both object and scene, respectively). This experiment indicated

that the visual content structure “pyramid” is theoretically robust, and it can be used to guide the

indexing process and classify the image labels manually and automatically. It is a systematic

category methodology for image indexing.

Although textual annotation of image retrieval is employed commonly, manual annotations are

expensive and waste lots of time and money. Sometimes, web-based image search engines are

always mismatch the annotation with the content of images, the result of querying for those

mismatch images would cluttered with irrelevant data. Furthermore, it is not clear how to select

images through noisy data sets, such as mislabel (Ben-Haim et al, 2006). Therefore, scholars try

to find out if it is possible to automatically annotate images and multimedia, or search images

without annotations. In Villa’s research (2010), they found an algorithm to allow users to search

for images through an intermediate database, which means a user can use query term to find out

a visual example from intermediate database, and then example would be used to search for the

images lack of annotation. This method can solve the problem stated before, when an image is

lack of label or mislabelled, it can still be found in this way.

2.3.2 Content based image retrieval

Furthermore there is another way for image search, which called content based image retrieval

(CBIR). It is for users to query through uploading image files and match similar images as search

result, and this way is also used for image search system (Smeulder et al, 2000). Content-based

image retrieval required computing the appropriate image decomposition and index, and storing

this extracted information for searches based on image content (Sclaroff 1997). This technique

need to build based on a huge image library to analyse multimedia content, however, thanks to

the development of inexpensive hardware and software for image acquisition, storage and

distribution, this encourage an increasing volume of digital images unloaded online to foster the

growth for CBIR (Acharya, 2011). In Acharya’s research they also mentioned that CBIR search

engines can translate users’ query image and search the similar images from the database in

some extent. Nevertheless, CBIR always forced user to describe their requirements with

uploading example images, it would be difficult to find result, if no visual example of the

information need is available (Villa et al, 2010). In order to solve the problem of users uploading

Page 13: People’s preferences, perceptions and frequency of using

13

picture as query terms, many research tried to develop new CBIR models. In Ben-Haim et al

research (2006) they propose an approach named Re-ranking sets of pictures by Exploiting

Consistency. This method is a hybrid of content-based retrieval and text-based retrieval. In this

method, users can retrieve with keyword and then clusters the results based on extracted image

feature. Another new model about content-based image search is called TSI-pLSA, developed by

Fergus et al in 2005. This method can handle the high intra-class variability and large proportion

of unrelated images returned by search engines.

2.3.3 Limitations of image search

General to say, there exist some problems related to image search, such as irrelevant images,

unclear images and unlabelled images (Ben-Haim et al, 2006). Furthermore user will find difficult

in obtaining the needed images, and developer will find difficult in images annotated

automatically (Covey 2002), or user needs are not taken into account appropriately (Burger

2010). Although Some scholar improved the image search system to handle the high intra-class

variability and large proportion of unrelated images returned by search engines (Fergus et al

2005), it is still difficult to meet individual users requirement according to their search

behaviours, as they might have different image needs in image search (Jorgensen and Jorgensen,

2005). In that way, there are still some limitations of image search, no matter using textual

annotation retrieval or content-based retrieval. Based on these limitations of image retrieval,

some researches were conducted. In Zhang et al study (2006), they root their hypothesis into

users’ behaviours, that customers would be easier to select terms generated by the image search

system rather than describing their target images by typing queries or uploading example images.

Therefore, they developed a method called interactive text-based image retrieval. Also new

technology such as organize search results (Rodden et al, 2001) and automatically annotate

images (Kennedy et al, 2009) should be used to increase the veracity of image search to meet

customers’ image seeking needs.

2.4 Image search behaviour

According to Zhang et al research (2006) about a commercial search engine queries logs, it shows

that more than 20% of image search queries can be categorized as location, nature, and daily life

related requirement. Therefore, image retrieval developed rapidly related to customers’ daily

lives and distinctive needs. Different individuals have different requirements in image retrieval

and the main challenge is an image can convey different meanings for users (Shatford-Layne,

1994). In order to improve image search veracity, researchers cannot only focus on the methods

and algorithms, but study users’ search behaviours as well. Therefore, in order to meet

Page 14: People’s preferences, perceptions and frequency of using

14

customers’ needs in image search, there are some aspects should be clarified. According to the

book Image and video retrieval, image search behaviour is a complicated interaction involves a

lot of factors, one method to search individuals’ behaviour is that examine and analyse the

transition made by individual user through one search query to another (Goodrum et al 2003).

However, this might be too complicated to study every single user’s search behaviour. According

to Jorgensen and Jorgensen (2005), because there are mass of factors relevant with image

content, images display and images search results would be categorized differently by their

qualities, and users’ visual perception might be various depending on those factors. Basically,

those factors could be reflected directly from image content, such as the size and colour of

images (Jorgensen 2005), therefore those factors are crucial for image retrieval. Furthermore,

search moves, traits and strategies which employed by users are also crucial when studying users’

behaviors through image seeking (Goodrum et al, 2003). As people usually consider some basic

factors when search images, for example images’ colour, sharp, hue, size and mood, those factors

might influence the similarity between image contents and individual’s expectation. Therefore, it

is necessary for developers build image retrieval system to consider those factors which influence

people’s behaviours on image searching (Burger, 2010).

2.4.1 Professional users’ search behaviours

Most recent researches are focus on professional users’ search behaviours. In Burger’s report

(2010), there are six factors affect reuse multimedia search, including images. For those arts

professionals who want to reuse media objects, they use keywords or classification information

in high frequency. Among six factors, Practices and feedback, visual and compositional features

are most important factors affecting search behaviours. Another search (Westman et al, 2008)

indicated when searching for an image, journalism professionals and non-professionals would be

affected by task type. Users always choose to combine content-based and textual search models

together to find visually tasks. On the other hand, conceptual tasks only required large number of

text queries and categories. Therefore, different classified image tasks need different search

methods. In another research, Fukumoto (2006) observed and tested undergraduate students’

search behaviours that were given certain image retrieval tasks. A popular search strategy was

found in this research: At first, student uses one or two keywords as query term to search for

images, and then the returned images were viewed. Finally the “back to home” or “previous

page” button was used to begin another search. According to this result, the author proposed a

new interface as “quick-browse” for an image retrieval system.

Page 15: People’s preferences, perceptions and frequency of using

15

2.4.2 Search query affects peoples’ search behaviours

Choi (2010) analysed users’ queries variation for image searches on the web. In the article,

study’s finding showed that query iterations varied with different task goals, job background, and

search expertise. Furthermore, the length of query would also vary with different image contents

and correlated with search engine operations. Jorgensen and Jorgensen (2005) reported an

analysis of search logs from a commercial image search web. In their report, users’ query

modification strategies were analysed and they made a conclusion that Boolean searching

seemed convenient but turn out ineffective and lead to query modification. Besides, descriptive

and thematic queries are more and more applied as query terms in image search retrievals. In

Pu’s research (2008) the failure of queries chose for web image search was analysed. The report

showed that queries would fail if it had high specificity or expertise, especially for some emotion

involved responses or abstract and conceptual refinements that required intellectual

interpretation (e.g., topic, genre, geographic, role, story, event, emotive, and abstract). In Chung

and Yoon’s research (2009), they analysed the characteristics of image search queries, especially

compared image attributes (abstract, generic, and specific) with basic metadata. The results

demonstrated that generic category appeared most frequently, users would use specific and

abstract attributes to search for images less frequently than generic attribute, and the colour

category was used least frequently.

Eakins, Briggs, and Burford (2004) made an experiment of images retrievals users, tried to find

out the requirements of users when search for images. The results demonstrated that

participants with high level of experience of searching for images would like to show high level

demand queries as search strategies rather than lower level features. In addition, the participants

were more likely to use keywords as query terms rather than menu choice or uploading example

image to search for an image. As the result, the authors made a conclusion that current CBIR

systems are inadequate, especially the interfaces part, which cannot satisfy user needs when

users cannot use keyword without example to search for images. Therefore it is the time to

develop new query paradigms. Yoon (2011) analysed user behaviour as the proliferation of

images in daily life. In this research, most participant chose Google or Google Image as search

engines to search for images, because they are familiar to users and have a user-friendly

interface. Although study finding indicated a high success rate of image retrieval, users still

wanted search engine could return more relevant images rather than irrelevant. Furthermore,

the search also found that student search images for various purposes, such as wallpaper of PC,

prepare for presentations, travel destinations and so on. This article analyse people’s needs of

image retrieval in daily life, but does not analyse how people operated image search engines.

Page 16: People’s preferences, perceptions and frequency of using

16

2.4.3 Similarity of images affects peoples’ search behaviours

In order to understand people’s behaviour on image search, it must be rooted in the

understanding of factors which influence peoples’ action through seeking images. In this way,

some scholar became to think the image’s similarity, whether it is a factor that influences

people’s search strategies, and if it would help to understand why the unrelated image search

result ratio is high at some extent. Little is known about how humans assess the similarity in the

case of complex images (Tirilly et al, 2012). However, one definition of two images similarity can

be found that if they are images of the same object or scene viewed under different imaging

conditions, those two images are similar (Jegou et al, 2008), or if they describe objects or scenes

from the same category and scale, those two images are similar (Fergus et al, 2005). So that if

customer find unsatisfied with the image search results, it would be necessary to check if the

concept of similarity is common to the different subjects of the search engine (Jorgensen et al

2001). To respect to an image query with low unrelated results, measures of image similarity are

needed to apply (Jegou et al, 2008), image retrieval systems aim at providing relevant images to

their users, but are actually mostly relying on the notion of image similarity (Tirilly et al, 2012).

Indeed, measures of image similarity are important to use in image retrieval system in order to

respect to an image query.

2.5 Conclusion

In conclusion, these sources provide a complete background about image retrievals, the types

and paradigms of image retrievals and some professionals and non-professionals behaviours of

image retrieval, and the potential issues of image retrievals. Furthermore, it can be concluded

that most researchers focus on professional users’ search strategies of image retrieval, but for a

commercial search website, they may be more interested in normal users’ needs of image

retrievals. Therefore, it is important to carry out this research. Users’ backgrounds and habits

may affect their search strategies, and some image search behaviour may originate from a user’s

daily life, job task, or academic life. These underlying factors affecting search strategies have not

been analysed yet. Based on above articles and reports, normal people’s search strategies should

be analysed to find out the way how to improve image search paradigms.

Page 17: People’s preferences, perceptions and frequency of using

17

3. Methodology

3.1 Introduction

In this research, quantitative methodology will be mainly applied to make a survey of

questionnaire, and some questions would use qualitative methods to analysis. In the survey, the

quantitative questions and Likert Scale (Hodge, 2005) will be applied to investigate people’s use

of image search in daily life and academic life. Critical Incident Technique (Flanagan, 1954) will be

also applied in this research, to investigate users’ most recently experience of searching for

images. All methodologies are used to understand people’s experience of carrying out image

searches, including people’s preferences, perceptions and frequency of use of image search

engines.

3.2 Quantitative methodology

During the late 19th and throughout 20th century, quantitative researches have been used to

study in many fields, such as psychology, economics, sociology, history, physics etc. Among those

quantitative strategies of search, these are survey, experiments, the less rigorous experiments

called quasi-experiments, correlational studies (Campbell & Stanley, 1963) and specific single-

subject experiments (Neuman & McCormick, 1995), which invoked positivist world view. This

research uses survey as a main methodology because it studies a sample of participants with a

series of questions and provides a quantitative or numeric description of trends, attitudes, or

opinions of a certain population. In this research, the data of people’s behaviours of image search

cannot be collected from every user of search engines; therefore, study a sample of people

becomes important for this research. Survey is just the way to study a sample of people, so it will

be used as a main methodology in this research. Besides, survey methodology applies principles

to the sample participants in the following steps. Firstly, design the relevant questions, and then

collect the data from participant as analysing sample, finally, process data and analysis data.

Although survey methodology is a quantitative technique, it still has errors and bias when

analysing statistics.

3.3 Qualitative methodology

In order to cover the errors or bias of survey, some qualitative method should be used. In

qualitative research, it tends to collect data directly from participants in the field of studying

issues. In the survey, in order to study the user’ perception of image search, some qualitative

Page 18: People’s preferences, perceptions and frequency of using

18

questions would be asked directly to investigate the objects RQ1.1 and RQ1.2 (seen in the

Introduction Chapter). Qualitative method observes or interviews participants directly and

studies participant experiences about the issue or problem, therefore, it will be fit for

investigating questions related to people’s perception. Furthermore, Qualitative method is not

similar as quantitative one. Participants do not need to fill questionnaire forms or follow

instruments of experiments. Mostly participants would interact with the researchers to collect

data (Creswell, 2009). Using qualitative method as early stage analysis with literature reviews to

make sure the research issue will help to carry out questionnaire survey as main methodology

and can cover some bias of quantitative method.

3.2 Critical incident Technique

Another method called the Critical Incident Technique (CIT) is used in this research to assist with

quantitative method. CIT studies human behaviours by asking them describing a set of

procedures in a certain “incident” (Flanagan, 1954). Critical incidents can be gathered in several

ways, normally participants are asked to describe their experience about the “incident”. The

means of descriptions can be writing down their experience in their own words, recording their

interviews or answering some certain questions of the “incident”. In my research, the CIT is used

as part of the questionnaire with several questions, to investigate RQ3.1, RQ3.2 and RQ3.3 (seen

in the Introduction Chapter). Those research questions are mainly about people’s most recent

experience of search for images and the experience would be a kind of incident at some extent. A

critical incident can make a significant contribution to the search issue. These questions are then

kept track of the “incidents” and used the “incident” data to solve relevant problems. Therefore,

CIT will be useful to investigate people’s behaviours on the most recent image search, and also

helpful to analysis underlying “incidents” of searching for images.

Recognising that all methodologies are imperfect and have limitations, I combined quantitative

and qualitative methods together to reduce bias as many as possible in any single method, and

wish to neutralize or cancel the biases of other methods. Mixed methods strategies are not

famous as quantitative or qualitative methodology but more practical in real life researches. In

1959, Campbell and Fiske developed the new term called multimethods to describe mixing

different methods together. They used multimethods to research in psychology field, especially

for the validity of psychological traits. After the success of utilizing their mixing methods, they

encourage other researchers to employ their multi-method matrix to study whether multi-

methods can be used to collect data in other fields. Therefore, I used quantitative method to

make questionnaire and combine scale, critical incident methodology together to carry out this

Page 19: People’s preferences, perceptions and frequency of using

19

research. Furthermore, using mixed methods can also helped to analysis different data collected

from samples.

3.3 Survey design

3.3.1 Questionnaire questions Q1-Q10

In this research, several questions were investigated, and overall aim of this research is to find

out the reason why people use search engine to search for images and how people behave when

using it. In the introduction Chapter, 3 main sections and 7 research questions of object were

designed in questionnaire with 19 questions. Besides, the questionnaire was designed based on

Yoon’s research called searching images in daily life. In his research, he focused on people using

search engines to search for images in daily life, and how they respond when searching for

images.

Firstly, the questionnaire was split into three parts. Part 1 included Q1-Q4 (Seen in Appendix A),

and those four questions were used to investigate the participants’ demographic, including age,

gender and their education background. In order to analysis simply, all the questions in this part

are close questions and design based on general survey. Then, in part 2, there are 6 questions.

The question number is Q5-Q10. (Q5. On average, how often do you use search engines to search

for images? Q6. How many times per week do you use search engine to search for images in your

job or study? Q7. How many times per week do you use search engine to search for images

OUTSIDE of your job or study?) Those questions were designed based the RQ 1.1 to investigate

participants’ frequency of using search engines to search for images. After that, Q8 (How

important is image search to you for the following tasks?) was designed to investigate the reason

why people used search engines based on RQ 1.1. Q9 (Which search engines do you normally use

to search for images?) and Q10 (Which of the following are most important to you when

conducing image searchers?) were designed based on RQ 1.2 to investigate the most popular

search engine and the reasons. Finally, an open question is offered for participant to list more

reasons. In those two parts, Q1 to Q10 were mainly designed to solve RQ 1.1 and RQ 1.2.

3.3.1 Questionnaire questions Q11-Q19

In part 3, Critical Incident Technique and Likert Scale are applied to investigate participants’ most

recent experience of searching for an image. Q11 (When was your most recent search for an

image) was designed to begin the investigation of search “incident” and this is a precondition to

investigate participant’s “incident”. Q12 (Why did you carry out this search) and Q13 (What was

Page 20: People’s preferences, perceptions and frequency of using

20

the topics on which you searched?) were asked the reason that they carried out this search and

what kind of topic of image did them search for. Another key issue need to investigate is which

search engine they used in their most recent image search (Q14 what was the search engine you

used for this search) and how easy they operated this search engines (Q15 How difficult was this

search, to find relevant images?). As reasons why people use image search engines would be a

personal question various with individual, utilizing Likert Scale can measure this perception in a

quantitative traits. Likert Scale is a quantitative methodology to measure people’s perception of

using image retrieval, so this method is used to ask about how people feel when operating search

engines. Besides, Q16 was designed to find participants felt successful or not for this image

search. If they felt successful, they should rank the scale of the reasons in Q17 (Please rate the

following statements, if you felt you were not successful in your image search (Jump to Q19)),

and if they felt unsuccessful, they also need to rank the reasons in Q18 (18. Please rate the

following statements, if you felt you were successful in your image search) of the questionnaire.

Finally Q19 (Please select ways the image search engine you used can be improved) was designed

to know which aspects of search engines should be improved and it will helpful for the future

study in this area. Questions in part 3 were mainly designed based on the RQ 2.1, 2.2, 3.1, 3.2,

3.3. Those questions were focus on their perceptions after search engine giving the results. The

results of this questionnaire would be important for building image search system. The similarity

of images could influence people’s need and the degree of satisfaction with the results after

searching certain types of image.

3.4 Practical

3.4.1 Participant

In order to carry out this research, the staff and students of the University of Sheffield became

the underlying participants. University students and staff may be a specific group in searching

image. In assumption, they would use image retrieval both in personal way and academic way.

This research will focus on the group of students and staff in the University of Sheffield to study

their behaviour on image retrieval and to prove above assumption. As a student or staff in the

university, they may need to search for images in different way and for different needs, especially

they can search for both job/academic needs and normal life needs. However, there will be some

special reasons of searching for images listed by students or staff in the university, and those

reasons may not be widely accepted. Nevertheless, choosing staff and students in the university

as participants can fit for this questionnaire, especially for investigating the academic/job needs

of image search.

Page 21: People’s preferences, perceptions and frequency of using

21

3.4.2 Ethical Aspects

In this research, no sensitive or confidentiality questions are contained. However, it directly

involves people in the research activities. As research involves human participant, ethical

problems are inevitable (Silverman, 2011). In this case, questionnaire and scale do not contain

sensitive questions, so the research would be low-risk. Furthermore letter of approval (Seen in

Appendix D) will be attached with questionnaire sent by email, and all the research questions are

anonymous. Every participant will be informed the purpose of this research (Seen in Appendix B),

the person undertaking and sponsoring the project, the potential risks that may have in this

research and potential benefits that may result. In this way, the whole research would undergo a

low risk to announce.

3.4.3 Pilot test

After finished the draft of the questionnaire, there existed 17 questions, and it is necessary to

take a pilot test which means to send the draft questionnaire with the ethics approval to a few

participants and let them fill it and find out whether it is practical or not. Therefore, before

collecting data from those participants, two people were invited to take the pilot test. During this

test, two things are aimed to achieve. One is to check the way of express in the questionnaire,

and to check whether the questionnaire convey clearly or not. Because the questionnaire might

have some ambiguous description of image search, participant might misunderstand some

questions. Another aim of pilot test is to check the structure of the questions, whether it could

use to collect the precise and sufficient data from the participants.

After taking the pilot test, these two people found that the question asked the frequency of using

search engines was unclear and easy to misunderstanding, especially the options of this question.

Although three question were designed to ask frequency of using search engines, those

questions’ options were different. After pilot test, those options were unified as the same pattern

to ask the frequency and it became easy to answer in the final questionnaire. In the draft

questionnaire, it did not contain the question of searching topics and difficulty, but the topic will

affect the difficulty of finding images. Therefore, the topics of last image search and how

difficulty participants felt when operating the search engines were added into the questionnaire

in order to investigate the difficulty of image search (Q14 and 15). In the draft questionnaire, Q17

and Q18 once were open questions, which were designed to ask whether participants feel

successful or not when carried out last image search and participants should also write down

their own reasons in those two questions. In the pilot test, participants mentioned that those

questions would be difficult to answer and people may lose their patient when answering such

Page 22: People’s preferences, perceptions and frequency of using

22

questions. Therefore, those questions was modified after pilot test, especially listed the possible

reasons for successful or unsuccessful image searches. After pilot test and modification, the

questionnaire contained 19 questions in the end.

3.4.3 Questionnaire distribution

The questionnaire was designed and sent out under the guide of the research in University of

Sheffield and obeyed the ethics approval. The questionnaire contains 3 parts with 19 questions,

and sent out by email through the University of Sheffield mail system and in the section of

“student volunteer”. With this survey email, ethics approval and information consent (Seen in

Appendix C) about this research was attached and sent to participants. Participants can read the

information consent to understand the aims of the research. This research used Google Docs to

create an online survey of questionnaire last for 15 days, since 15/07/2013 to 31/07/13. The

survey email was sent to all the students and staff in the University of Sheffield and finally 84

people in the University of Sheffield participated in this survey.

Page 23: People’s preferences, perceptions and frequency of using

23

4. Result

4.1 Demographic information Q1-Q4

In this research, the survey was divided into 3 parts, part one is about participants’ demographic

information, part two is about people’s perceptions of searching for images and part three is

about people’s most recent experience of searching for images. In the survey, 84 participants

completed the survey. Among them, 38.1% were male and 61.9% were female (Table 1). The

number of female participants was approximately 1.5 times the size of male participants. 58.3%

of the participants’ age ranged from 18 to 24, no matter the participant was male or female.

Nearly 1 out of 5 male participants’ age ranged from 18 to 24 and 38.1% female participants aged

18 to 24. In different age range, in the range of 25-30, gender rate was 1.5 (female: male),

however, in the range of 31-40, male participants were more than female. In Table 2, 86.9% of

the participants were students, only 13.1% of participants were employed. Among students,

42.9% were Master student, 28.6% were undergraduate student, and 6.0% were doctoral

student. However most employed were undergraduates.

Table 1 Q1-Q2 gender and age scale of people who participated in the survey

Age Total

18-24 25-30 31-40 41-50 Over 50

gender

Male Count 17 7 7 0 1 32

% of Total 20.2% 8.3% 8.3% 0.0% 1.2% 38.1%

Female Count 32 13 5 1 1 52

% of Total 38.1% 15.5% 6.0% 1.2% 1.2% 61.9%

Total Count 49 20 12 1 2 84

% of Total 58.3% 23.8% 14.3% 1.2% 2.4% 100.0%

Page 24: People’s preferences, perceptions and frequency of using

24

4.2 Perceptions regarding searching for images Q5-Q10

4.2.1 Q5-Q7 frequency of searching for images

In the part two of the questionnaire, participants were asked to describe the frequency of using

image search systems to search for images. Three different frequencies of searching for images

were asked to describe, and 4 scales were used to analysis (Table 3). In the scale, 1-5 represent “1

to 5 times per week”, 5-10 represent “5 to 10 times per week”, 10-20 represent “10 to 20 times

per week” and >20 represent “more than 20 times per week”. Furthermore, Frequency 1 means

on average, how often they used search engines to search for images. In this question, 34.5%

Table 2 Q3-Q4 Education background and occupation of participants

occupation Total

Employed Student

education

Doctorate

Count 0 5 5

% within

education

0.0% 100.0% 100.0%

Further education

Count 0 2 2

% within

education

0.0% 100.0% 100.0%

High school

Count 0 2 2

% within

education

0.0% 100.0% 100.0%

Master

Count 3 36 39

% within

education

7.7% 92.3% 100.0%

PG dip

Count 1 0 1

% within

education

100.0% 0.0% 100.0%

Professional

Count 1 4 5

% within

education

20.0% 80.0% 100.0%

Undergraduate

Count 6 24 30

% within

education

20.0% 80.0% 100.0%

Total

Count 11 73 84

% within

education

13.1% 86.9% 100.0%

Page 25: People’s preferences, perceptions and frequency of using

25

people believed they searched for images 1 to 5 times per week on average, 27.4% participants

searched for images 5 to 10 times per week, 21.4% participants searched for 10 to20 times and

even 16.7% participants searched more than 20 times per week. Frequency 2 is the frequency of

searching for images in job/study, and 59.5% participants searched for 1 to 5 times per week for

job/academic needs. Frequency 3 is the frequency outside job/study. The participants described

nearly the same frequency in job/study and outside, 48.8% participants searched outside job 1 to

5 times per week, but outside job/study was in higher frequency and scatter wider than in

job/study.

Table 3 Q5-Q7 Three different frequencies of searching for images. Frequency 1 was on average image search

frequency, 2 was search for images in job/study, and 3 was search for images outside job/study. 1-5 means 1-5

times per week, 5-10 means 5-10 times per week, 10-20 means 10–20 times per week, and >20 means more than

20 times per week.

Valid

1 - 5 5 - 10 10 - 20 >20 Total

Frequency 1 29 23 18 14 84

Percent 34.5% 27.4% 21.4% 16.7% 100.0%

Frequency 2 50 24 7 3 84

Percent 59.5% 28.6% 8.3% 3.6% 100.0%

Frequency 3 41 23 11 9 84

Percent 48.8% 27.4% 13.1% 10.7% 100.0%

4.2.2 Q8 the reason why carry out image search

The reason why they carried out image search was asked after the frequency of using search

engine to search for images. Different participants gave the different answers. The survey list 8

reasons for participants to rank (1 is very unimportant, 5 is very important). Most participants

thought that finding out what something looks like was the most important reason among all

options to search for images (Table 4). Then curiosity ranked No.2 among those reasons but less

important than finding out what something looks like. People also search for image in need of

preparing presentation and finding out travel destination. Verifying names and getting the idea of

purchasing were also the important reasons why people carried out an image search. For the

reason finding beauty/fashion items, it varied widely between participants, and the numbers of

the 5 scales chosen by participants were primarily equal. Besides those listed reason, participants

also wrote down other reasons for image search. One participant wrote that searching for images

to enhancing email communication and enriching class hand-outs. Another participant wrote that

finding out original image and photograph was important especially those images were reposted

Page 26: People’s preferences, perceptions and frequency of using

26

too many times. Two participants wrote that sharing amusing memes with others was the reason

to search for images, which means replying others on the internet with small images and photos

to enhance your ideas. There are still some other reasons for conducting an image search, such

as: finding out porn, program or apple icons, adding to blogs etc. Therefore the further discussion

of this question would be presented in the next section.

Table 4 Q8 Why carry out image search. The table shows the reasons why carried out image search, and

used Likert Scale to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant, and 5

means very important. In the table, Range means how many numbers participants used to measure the

reasons.

Range Mean Std. Deviation

Something looks like 3 4.50 .720

Curiosity 4 3.99 .814

Prepare presentation 4 3.93 1.159

Travel destination 4 3.79 1.233

Verify names 4 3.48 1.156

Ideas for purchasing 4 3.35 1.187

Beauty/Fashion items 4 3.10 1.314

Computer wallpaper 4 2.51 1.266

Valid N (listwise)

4.2.3 Q9-Q10 which search engine is popular and why

In part two, participants were also asked to describe which search engine systems they usually

used for image search (Figure 1). Q9 was a multiple choices question, and participants can

choose as many options as they wanted. Furthermore, participants can choose search engines

even they only used once. As the results, the majority of participants used Google as their

favourite search engines, and using search by image (73, 87.0%) more than web search system

(64, 76.2%) to search for images. After that, Bing was ranked No.3, and 10.7% participants chose

it as their usually used image search engines. Then, 9.5% participants used Baidu to search for

images, which is a Chinese search engine and 2.4% participants chose Tineye, which is a reverse

image search engine in professional image search. 2.4% chose Flickr, which is image hosting

website as image search engine, 1.2% participant chose Ask, which is an American search system

and 1.2% participant used Yandex, which is a Russian search engine. Participants were also asked

the reason why they chose those search engines to search for images (Table 5). After Q9, some

reasons were listed for participant to rank (1 is very unimportant, 5 is very important) in Q10. The

highest ranked reason to choose a search engine is that the chosen search engine would return

Page 27: People’s preferences, perceptions and frequency of using

27

good search results based on participants’ search query. This reason ranked 4.69 was higher than

the second reason which was the familiarity with search engines. Then the user-friendly interface

and returning large numbers of various images were two important factors for choosing search

engines to search for images. Offering relevant text with images, Availability of copyright-

unrestricted images and sharing images with others were all ranked above 3 but less than 4 in the

survey as factor for choosing search engines.

Table 5 Q10 the ranking of reasons for choosing image search. In the table, reasons for choosing search

engines were listed, and used Likert Scale to rank those reasons. In the scale, 1 to 5 were used, 1 means very

unimportant, and 5 means very important. In the table, Range means how many numbers participants used

to measure the reasons.

Range Mean Std. Deviation

Good results 4 4.69 .711

Familiarity 3 4.35 .736

Interface 4 4.43 .749

Image amount 4 4.07 1.039

Relevant text 4 3.81 1.081

Copyright 4 3.37 1.138

Ability to share 4 3.15 1.256

Valid N (listwise)

Figure 1 Q9 Popular searcher engine to search for images (multiple choices can be made)

73 87.0%

Page 28: People’s preferences, perceptions and frequency of using

28

4.3 Most recent experience of searching for images

4.3.1 Q11 when participant carried out the last image search.

After analysing above questions, the most recent experience of searching for images of

participants were asked. From question 11 to question 19 were all about this experience. The

question 11 was asked when participant last time carried out image search (Table 6). Most

participants (38.1%) thought they carried out the most recent image search 1-3 days ago. Then

about 42.9% participants carried out image search within a day in total. Nearly 10% thought they

carried out an image search more than 4 days ago but within a week. Therefore, primarily 91.7%

participants carried out a search for images within a week, and only 8.3% participants carried out

an image search more than one week ago.

Table 6 Q11 When was the Last time participants searched for images.

Frequency Percent Cumulative Percent

Valid

Within an hour 6 7.1% 7.1%

1-2 hours ago 10 11.9% 19.0%

More than 3 hours ago but

within a day

20 23.8% 42.9%

1-3 days ago 32 38.1% 81.0%

More than 4 days ago but

within a week

9 10.7% 91.7%

More than one week ago 7 8.3% 100.0%

Total 84 100.0%

4.3.2 Q12-Q13 topics and reasons carrying out last image search

In Q12 and 13, the topics and the reasons to carry out last image search were asked (Table 7,

Table 8, and Table 9). Nearly 52.4% people carried out last image search for curiosity or

entertainment. About 19% participants carried out image search for job/academic needs, and

37.5% participants who searched for job/academic needs searched medical related material or

plants/animals related images last time. In the reason of gaining knowledge, most participants

(27.3%) select art related materials as search topics, and in the reason of problem solving, 75.0%

participants searched for travel places/destination. 16.7% of all the participants chose figure of

people as search topics and they always carried out either for curiosity or for entertainment

reason. Different participants had different search topic. In the survey, 15 possible topics were

listed and every topic had its own chooser. There were still 8.3% participants chose others option

in Q12. Therefore, the topics and reasons for carrying out image search varied with different

Page 29: People’s preferences, perceptions and frequency of using

29

participants.

Table 7 Q12 Topics participants chose to search for images in their most recent image search

Frequency Percent Cumulative Percent

Valid

News 3 3.6% 3.6%

Plants/Animals 6 7.1% 10.7%

Cook/Food related material 7 8.3% 19.0%

Fashion/Beauty items 3 3.6% 22.6%

Arts related materials 6 7.1% 29.8%

Sports related material 3 3.6% 33.3%

Places 12 14.3% 47.6%

Mechanism related material 1 1.2% 48.8%

Electronics products 3 3.6% 52.4%

Information technology related material 3 3.6% 56.0%

Medical related material 8 9.5% 65.5%

Films, Books or Music related material 8 9.5% 75.0%

Figure of People 14 16.7% 91.7%

Others 7 8.3% 100.0%

Total 84 100.0%

Table 8 Q13 Reasons participants chose to search for images in their most recent image search

Frequency Percent Cumulative Percent

Valid

Curiosity 24 28.6% 28.6%

Entertainment 20 23.8% 52.4%

Job/ academic needs 16 19.0% 71.4%

Gain knowledge 11 13.1% 84.5%

Problem solves 8 9.5% 94.0%

others 5 6.0% 100.0%

Total 84 100.0

Page 30: People’s preferences, perceptions and frequency of using

30

Table 9 Q12-Q13 correlation of topics and reason participant chose to carry out last image search. In the table,

Frequency was used to analysis the correlation of topics and reasons.

Reason Total

Curiosity Entertainment Job/ academic

needs

Gain

knowledge

Problem

solves

others

topics

News 1 0 1 1 0 0 3

Plants/Animals 0 1 3 1 0 1 6

Cook/Food related

material

3 1 0 2 0 1 7

Fashion/Beauty items 1 0 0 1 0 1 3

Arts related materials 0 2 1 3 0 0 6

Sports related material 1 2 0 0 0 0 3

Places 4 2 0 0 6 0 12

Mechanism related

material

0 0 0 1 0 0 1

Electronics products 0 2 0 1 0 0 3

Information technology

related material

0 0 2 0 1 0 3

Medical related material 1 0 6 0 0 1 8

Films, Books or Music

related material

4 3 0 1 0 0 8

Figure of People 8 6 0 0 0 0 14

Others 1 1 3 0 1 1 7

Total 24 20 16 11 8 5 84

4.3.3 Q14-Q16 search engine and difficulty & satisfaction of search engines

When asked which search engine the participant used for last image search, the majority of

participants (96.4%) used Google as the search engine, no matter web search or search by image

(Table 10). Only 1 (N=84) participant used Yahoo!, 1 (N=84) participant used Baidu (Chinese), and

1 (N=84) participant used Bing to carry out most recent image search. 63.1% participants thought

using search engines to search for images was easy to operate (Figure 2), and 90.5% participants

thought they were satisfied with the search results, and felt successful with this search (Figure 3).

The majority people (61.9%) believed that Google web search was easy to operate and satisfied

with its search result. In this survey, only 9.5% participants felt unhappy with this search. Among

those participants, 37.5% participants thought this image search was neither successful nor

Page 31: People’s preferences, perceptions and frequency of using

31

failure, and only 12.5% participants thought this search was very failure. Among those

participants who felt unsuccessful, 12.5% used Google search by image, 25.0% used Google web

search. Then 4.7% out of all participants believed this image search was failure and they all used

Google search by image and among them 75% people thought search by image was somewhat

difficult to operate and 25% thought it was very difficult. Finally, 1.2% participant used Baidu and

felt very difficult to operate and very unsatisfied with the search engine and cannot find out what

he/she searched.

Figure 2 Q15How difficult to operate search engine to search for images

Figure 3 Q16 Satisfaction of search engine to search for images

2.4% 4.8%

11.9%

17.8% 63.1%

50.0% 40.5%

3.6% 4.7%

1.2%

Page 32: People’s preferences, perceptions and frequency of using

32

Table 10 Q14-Q16 How participants satisfied with the search engine, how difficulty they felt to operate the

search engines, and which search engine were used to carry out the most recent image search

Satisfaction Difficulty Total

1

(Easy)

2 3 4 5

(Difficult)

Very successful, I

found exactly

what I required

Search

engine

Google (web search) 14 1 0 0 15

Google (search by image) 22 2 2 1 27

Total 36 3 2 1 42

Successful

Search

engine

Google (web search) 4 8 3 15

Google (search by image) 11 3 3 17

Yahoo! 0 1 0 1

Bing 1 0 0 1

Total 16 12 6 34

Neither

successful nor

failure

Search

engine

Google (web search) 0 2 2

Google (search by image) 1 0 1

Total 1 2 3

Failure

Search

engine Google (search by image)

3 1 4

Total 3 1 4

Very failure, I did

not find what I

required

Search

engine Baidu (Chinese)

1 1

Total 1 1

Total

Search

engine

Google (web search) 18 9 5 0 0 32

Google (search by image) 34 5 5 4 1 49

Yahoo! 0 1 0 0 0 1

Baidu (Chinese) 0 0 0 0 1 1

Bing 1 0 0 0 0 1

Total 53 15 10 4 2 84

4.3.3 Q17-Q18 successful and unsuccessful with the last image search

In Q17, after analysing the results of participants’ perception about most recent experience of

search for images, the reason why they felt successful or unsuccessful were asked. In the above

question, there 76 (90.5%) participants felt successful with their last time image search. However,

only 74 (88.1%) finished the reason ranking box about how they felt the search would be success

Page 33: People’s preferences, perceptions and frequency of using

33

(Table 11). The ranking box listed some reasons about being successful with the search results

and using 1-5 scale to rank. It was the same with the unsuccessful part. In the successful part,

most participant thought that the search engine had user-friendly interface and easy to operate

are the most important reasons to feel satisfy with the search results, and return relevant images

about the search query terms was also an important reason. Then, some participants believed

that image search engine in the native language would be a reason to choose for searching for

images, but some did not. Returned images with high resolution and copyrights seemed less

important for those participants to feel satisfy with the search engines.

In Q18, for the unsuccessful part (Table 12), although only 8 people ranked the reasons, there

were still some interesting findings. The reason that search engine returned irrelevant images as

results was the most important reason for the participants to feel unhappy with the image

search. Then, images with low resolution and the search engine unfamiliar to the searchers were

less important reason for them to feel unhappy. They did not think that search engine difficult to

operate and images with the wrong size were the important reasons to feel unsatisfied. No

matter for the participants satisfied with search engine or not, images with copyrights and

unlimited to use would not be the reason to conduct those feelings.

Table 11 Q17 the possible reasons for participants having a success experience in image search. In

the table, Likert Scale was used to rank those reasons. In the scale, 1 to 5 were used, 1 means very

unimportant, and 5 means very important.

N Range Mean Std. Deviation Variance

Statistic Statistic Statistic Std. Error Statistic Statistic

Easy to operate 74 2 4.86 .044 .382 .146

Quickly 74 1 4.80 .047 .405 .164

Relevant 74 2 4.61 .066 .569 .324

Native language 74 4 4.22 .147 1.264 1.596

High resolution 74 4 3.97 .110 .950 .903

With copyrights 74 4 3.07 .109 .941 .886

Valid N (listwise) 74

Page 34: People’s preferences, perceptions and frequency of using

34

Table 12 Q18 the possible reasons for participants having a failure experience in image search. In the

table, Likert Scale was used to rank those reasons. In the scale, 1 to 5 were used, 1 means very unimportant,

and 5 means very important.

N Range Mean Std. Deviation Variance

Statistic Statistic Statistic Std. Error Statistic Statistic

Irrelevant 8 2 4.00 .267 .756 .571

Low resolution 8 4 3.25 .526 1.488 2.214

Unfamiliarity 8 3 3.00 .423 1.195 1.429

No copyrights 8 3 2.88 .441 1.246 1.554

Wrong size 8 3 2.75 .453 1.282 1.643

Difficult to use 8 3 2.00 .423 1.195 1.429

Valid N (listwise) 8

4.3.3 Q19 aspects of search engine to improve

After research participants’ perceptions about the most recent search experience, they were

asked to select which aspects should be improved of the search engines and the options of this

question were multiple choose that participants can choose as many as they want. 66.7%

participants thought they wished those search engines could return more relevant images and

fewer irrelevant images, although 90.5% people felt satisfied with the search results of those

search engines. Then, 50% participants thought that search engines should improve the quality

or size of the returned images and provide related text information about those images. 45.2%

participants needed more related information about the images, and 28.6% participants thought

that providing more search options or improving the interface should be important for those

search engines. 23.8% participants thought search engines should improve the way that returned

images displayed, and the same number of people thought search engines should display the

copyrighted images or where those images came from. In the other options, only one person

chose the others option and thought that search engine was fine enough, no more place need to

improve.

Page 35: People’s preferences, perceptions and frequency of using

35

Table 13 Q19 In which aspects, search engines for image search can be improved (multiple choices can

be made)

Frequency Percent of

participant

Percent of

response

Cumulative

Percent

Valid

Returning more relevant images and

fewer irrelevant images

56 66.7 27.5 27.5

Improving the quality or size of the

returned images

42 50.0 20.6 48.0

Providing related information about the

image

38 45.2 18.6 78.4

Providing more search options or

improving the interface

24 28.6 11.8 59.8

Improving the way that returned images

are displayed

20 23.8 9.8 88.2

Displaying copyrighted images or images

from commercial websites

20 23.8 9.8 98.0

Others 4 4.7 2.0 100.0

Total 204 242.8 100.0

Page 36: People’s preferences, perceptions and frequency of using

36

5. Discussion

This research indicated University students and staff’s perceptions on using search engines to

search for images in job/study and outside. Most people participants in this survey were female,

because they may be more patient than male. Besides, this survey was carried out through

university mail system, so student participants were more than staff, and most students were

undergraduate students.

Firstly, in this study, different participants had their different frequency of image search relating

to the RQ1.1. Although most people would search for images 5-10 times per week, more than

16% participants thought they would search for image more than 20 times per week.

Furthermore, participants would like to search images 1-5 times per week for job/academic

needs and more or less the same times for personal reasons. However, the overall people chose

more times per week for personal reason search than job/academic needs. Although in their

most recent experience of image search, less than 20% people described that they searched

images for job/academic need, they all ranked job/academic need was an important reason for

them to search for images. In the survey, the most significant reason to carry out image search is

to find out what something looks like, and then is for curiosity. After that is the reason to prepare

presentation and find out the travel destination. In Yoon’s research (2011), finding images as

computer wallpaper was an important reason for image search, however, in this survey, most

people thought it was unimportant and only ranked 2.51 on average (1=very unimportant, 5=very

important). For the reason to find beauty/fashion items, the overall score was 3.10 on average,

but for male, the score was only 2.66 and for female, the score was 3.25. There exists huge

difference in this reason for image search. The study found that female felt finding

beauty/fashion items was more important than male.

Secondly, according to RQ1.2, 2.1 and 2.2, most people choose Google (web search) and Google

(search by image) as their most favourite search engines to search for images. The reason to

choose this search engine is that it would return good image results. After that, familiarity and

user-friendly interface were another two crucial reasons to choose this search engine to search

for images. Therefore, image searchers usually value the search results than other aspects of

search engine. A good search engine can satisfy user with good and relevant search results. This

would be important for search engine development. In the survey, participants still chose

returning more relevant images and fewer irrelevant images as an aspect that search engines

need to improve. Therefore, no matter how interface improved and large amount of images

returned, the most important respect need to improve is relevant image results. As most people

Page 37: People’s preferences, perceptions and frequency of using

37

choose Google as search engine and chose the reason why used it is that it can return good

search results. Therefore, Google may return more relevant images than other search engines so

that people were more likely to use it. In Fergus’s research, users can use Google’s image search

engine by querying key world, and Google would match key words with textual annotation to

return the results. While for Yahoo!, the same way was applied to its image search function, and

this made Yahoo! become the strongest competitors with Google in the world (Zhang et al, 2006).

However, in the survey, it was not hard to tell that only 6.0% participants choose Yahoo! as search

engine. Therefore, in some extent, Yahoo! cannot be the strongest competitor with Google.

Furthermore, Google no longer only used textual annotation for image search. It also applied

CBIR in “search by image” section for image search. In the survey, around 90% participants used

“search by image” function. Although they may misunderstand the meaning of “search by image”

and obscure this function with “image search”, they would still be a lot of people used “search by

image” function before. So in the survey, the most popular search engine to search for images

was Google.

Thirdly, CIT was applied to analyse the most recent experience of searching for images in order to

answer the RQ 3.1, 3.2. In Sudman & Bradburn’s study, time and other factors would affect

memory on response in surveys (1973). However, if the stimulate of memory were too strong,

and recalled by people several times, the memory would not fade but strengthen. In the survey,

participants were asked to answer when their most recent image search carried out, and more

than 80% participants carried out image search within 3 days. Therefore, the CIT of their most

recent image search would be at some extent reliable and accurate. In this study, most people

felt Google was easy to operate when searching for images, and Bing and Yahoo! were not hard

to operate. People who used Baidu (Chinese) found it was not easy to operate. Different people

carried out last image search for different reasons. Most people searched for curiosity or

entertainment reasons, and for this reason most participants chose figure of people as search

topic. Job/academic need and Gain knowledge were another two important reasons to carry out

last image search, and for those reasons, most people chose medical related images. In problem

solves reason, most people chose places as topics. They might believe that Google map was

another kind of image or they just simply searched the images showed the view of the place.

Fourthly, in the survey, most people felt successful with this image search. Only 8 people felt

unsuccessful with this image search (Table 10). In CIT, “incident” has its significant value to study,

so those 8 “incident” case would be analysis in details. In those “incident” cases, 2 people were

male and 6 were female, and the ratio of female to male was higher than overall ratio. Female

would feel more unsatisfied with the search results than male. After that, among those 8

Page 38: People’s preferences, perceptions and frequency of using

38

participants, 4 people felt failure and they all used “search by image” function. Among those 4

people, 3 participants thought “search by image” function was not easy to operate and 1

participant thought it was difficult to search. “Search by image” function is a CBIR search option,

and users should upload images as query terms to find similar images as search results. The

search process might be complex for some users, so they may felt difficult to operate. However,

because they may misunderstand the meaning of “search by image”, it could be that people felt

hard to operate image search function. Nevertheless, most people used Google (both web search

and search by image) for last image search and felt easy to operate the search engine, and the

image search interface was almost exactly the same with web search. Therefore, if those people

misunderstanding “search by image”, they may not feel hard to operate. Google search by image

would hard to operate and users did not satisfy with the search results. Besides, Tineye is also a

CBIR search engine, and users need to add images’ URL or uploaded image as search query.

However, no one used it as search engine for last image search. Therefore, CBIR might not be

widely used by people or not as popular as TBIR search engine. Furthermore, only one people

used Baidu (Chinese) to search for images in those 8 cases, felt difficult to operate the search

engine and did not satisfy with the search results. The reason for that was return irrelevant

images.

Finally, according to RQ3.3, no matter people feel satisfied with the search results or not, they all

wished search engines can return more relevant images and less irrelevant images. The reasons

for participants feel satisfy with the search engine is because it returned relevant images quickly

and in native language, and the reasons for not satisfaction is because it returned irrelevant

images and with low resolution. Therefore, people thought search engines need to improve its

search results. Furthermore, some people thought images should be up to date and with nice

quality or changeable size. Providing related information was also another aspect to improve.

However, people did not care whether the image displayed copyrights or not, not matter

participant satisfied with the search results or not.

Page 39: People’s preferences, perceptions and frequency of using

39

6. Conclusion

6.1 Summary

With the increase growth of image retrieval, it becomes significant to analysing people’s

perception on image retrieval. The frequency, perceptions and preferences of people searching

for images are very critical to develop search engines. In this research, 84 participants completed

the questionnaire, and all of them are students or staff in the University of Sheffield. The survey

was designed to find out people’s behaviours on the image search and which search engine was

the popular engine for image search according to research questions. In the questionnaire, part

one was designed to collect participants’ demographics. Most participants were undergraduate

and master students, and most of them were female.

In the research, part 2 was designed to collect basic information about participant using search

engines to search for images relating to RQ1.1 and RQ 1.2. In this part, the most popular search

engine was Google, which was used by 87.0% participants in the survey. On average, people

would use search engines to search for images 5-10 times per week. Participants used search

engines nearly the same times to search for images for job/academic needs and personal needs.

They usually searched for both needs 1-5 times per week. Most people use search engine to

search for images to prepare presentation, find out what something looks like, find travel

destination, verifying names etc. They carry out images search for curiosity, entertainment,

gaining knowledge, solving problems and job/academic need.

Finally, in part three, the questions were designed to analyse participants’ most recent

experience of image search and to answer the RQ 2.1, 2.2, 3.1, 3.2, 3.3. 90.5% participants

thought their last image search was successful and the search engine was easy to operate. Only

9.5% participants thought it was unsuccessful. Among them, most people used Google search by

image function and thought it was hard to operate. Participants would search some medical

related materials for job/academic needs, and search people’s figure for curiosity or

entertainment. The reason which people thought this image search was successful would be that

search engines returned relevant images quickly and was easy to operate. And the reason for

unsuccessful image search was that search engines returned irrelevant images or images with

low resolution. In the end, participants thought search engine need to return more relevant

images than irrelevant images. Search engines also need to improve the quality or size of the

returned images, provide related information about the image, and provide more search options

or improve the interface.

Page 40: People’s preferences, perceptions and frequency of using

40

6.2 Limitation

There are some limitations of this survey. First, the participants in this survey are only from the

University of Sheffield, the age, education background and other personal information may

restrict in a limited range. Furthermore, locational limitation is also a problem, as people around

world have different habit in searching for images, therefore the survey may not represent all the

phenomena in the image search. Although conducting pilot tests, there are still some problems

of misunderstanding or ambiguous terms. Because only two people took the pilot test, there may

exist some issued not be found by those two people. Finally, the survey is only displayed online,

so that it lacks the face to face communication. It is not so precise when studying participants’

behaviours through the survey, especially for some experience questions.

6.3 Future study

More researches should be taken in the CBIR search engine to find out why people thought it was

hard to operate and how to improve the search engines with these reasons. As this research

indicated, although most people satisfied with the search results, people still believed that search

engine should return more relevant results. Therefore, the algorithm of returning relevant image

as search results need to improve, and would be better to combine CBIR and text annotation

together to search for images. People prefer to use search engines that they are familiar with or

the image search interface is user-friendly. Hence, improve the search engine interface and make

it convenient to operate are very important in the future work.

(Word count: 11,772)

Page 41: People’s preferences, perceptions and frequency of using

41

Reference

Acharya, S., & Devi, M. R. V. (2011). Image retrieval based on visual attention model. Procedia

Engineering, 30(2011), 542–545.

Ben-Haim,N. Babenko B.and Belongie,S.(2006). ImprovingWeb-based Image Search via Content

Based Clustering. IEEE.

Burger,T.(2010). A Model of Relevance for Reuse-Driven Media Retrieval. Proceedings of the 2nd

Workshop on Semantic Multimedia Database Technologies, co-located with the 5th

International Conference on Semantic and Digital Media Technologies (SAMT).

Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the

multitrait-multimethod matrix. Psychological bulletin, 56(2), 81.

Campbell, D. T., Stanley, J. C., & Gage, N. L. (1963). Experimental and quasi-experimental

designs for research (pp. 171-246). Boston: Houghton Mifflin.

Creswell. J.W (2009). Research design: Qualitative, quantitative, and mixed methods

approaches. Sage.

Chen, H.-L., & Rasmussen, E.M. (1999). Intellectual access to images. Library Trends, 48(2), 291–

302.

Choi, Y., & Rasmussen, E. M. (2003). Searching for images: The analysis of users' queries for image

retrieval in American history. Journal of the American Society for Information Science and

Technology, 54, 498–511.

Covey, D.T. (2002). Usage and usability assessment: Library practices and concerns.

Washington, DC: Council on Library and Information Resources, Digital Library Federation.

Eakins, J. P., Briggs, P., & Burford, B. (2004). Image retrieval interfaces: A user perspective. In P.

Enser, Y. Kompatsiaris, N. E. O'Connor, A. F. Smeaton, & A. W. M. Smeulders (Eds.), Lecture notes

in computer science. Image and video retrieval, 3115. (pp. 628–637) Berlin, Germany: Springer-

Verlag.

Fergus, R., Fei-Fei, L., Perona, P., & Zisserman, a. (2005). Learning object categories from Google’s

image search. Tenth IEEE International Conference on Computer Vision (ICCV’05) Volume 1,

1816–1823 Vol. 2.

Page 42: People’s preferences, perceptions and frequency of using

42

Flanagan, J. C. (1954). The critical incident technique. Psychological bulletin,51(4), 327.

Fukumoto, T. (2006). An analysis of image retrieval behavior for metadata type image database.

Information Processing and Management, 42, 723–728.

Goodrum, A., Bejune, M. and Siochi, A. C. (2003). A state transition analysis of image search

patterns on the web. Image and Video Retrieval, 281-290

Hodge, D. R. & Gillespie, D. F. (2005). Phrase Completion Scales. Encyclopedia of Social Measurement, Vol3, 53–62. San Diego: Academic Press

Jegou, H., Douze, M. and Schmid, C.(2008) Hamming embedding and weak geometric consistency

for large scale image search. In Proceedings of the European Conference on Computer Vision

(ECCV), vol1, 304-317.

Jorgensen, C., Jaimes, A., Benitez, A. B., & Chang, S. F. (2001). A conceptual framework and

empirical research for classifying visual descriptors. Journal of the American Society for

Information Science and Technology, 52(11), 938-947

Jorgensen, C. and Jorgensen, P. (2005). Image querying by image professionals. Journal of the

American Society for Information Science and Technology, 56/12, 1346-1359.

Kennedy, L. Slaney,M. and Weinberger, K.(2009) Reliable tags using image similarity: mining

specicity and expertise from large-scale multimedia databases. In Proceedings of the 1st

workshop on Web-scale multimedia corpus (WSMC), pages 17-24, Beijing, China.

Neal, D. (2009), Introduction. Bulletin of the American for Society Information Science and

Technology, 35: 6–12.

Neuman, S. B., & McCormick, S. (1995). Single-Subject Experimental Research: Applications

for Literacy. Order Department, International Reading Association, 800 Barksdale Road, PO

Box 8139, Newark, DE 19714-8139 (Book No. 128: $11 members, $16 nonmembers)..

Pu, H. -T. (2008). An analysis of failed queries for web image retrieval. Journal of Information

Science, 34, 275–289.

Rieh, S. Y., and Hilligoss, B. (2008). College students' credibility judgments in the information-

seeking process. In M. J. Metzger, & A. J. Flanagin (Eds.), Digital media, youth, and credibility (pp.

49– 72). Cambridge, MA: MIT Press

Page 43: People’s preferences, perceptions and frequency of using

43

Rodden, K. Basalaj, W. Sinclair, D. and Wood, K. (2001) organisation by similarity assist image

browsing? In Proceedings of the ACM SIGCHI conference,Seattle, WA, United States.

Rui, Y., Huang, T. S. and Chang, S. (1999). Image Retrieval: Current Techniques, Promising

Directions, and Open Issues. Journal of Visual Communication and Image Representation,

10/1, 39-62.

Sudman, S., & Bradburn, N. M. (1973). Effects of time and memory factors on response in

surveys. Journal of the American Statistical Association, 68(344), 805-815.

Sclaroff, S., Taycher, L., & Cascia, M. La. (1997). ImageRover : A Content-Based Image Browser for

the World Wide Web 1 Introduction 2 Approach, 2–9.

Shatford-Layne, S. (1994). Some issues in the indexing of images. Journal of the American Society

for Information Science, 45(8), 583–588.

Silverman, D.(2011).Ethics and qualitative research. Qualitive research, 3rd edition, III,416-438.SAGE publication,Ltd.

Smeulder,A.W., Worring, M., Santini, S., Gupta, A., and Jain, R.(2000). Content-Based Image

Retrieval at the End of the Early Year. IEEE Trans.

Tirilly, P., Mu, X., Huang, C., Xie, I., Jeong, W., & Zhang, J.(2012). On the consistency and features

of image similarity. Proceedings of the 4th Information Interaction in Context Symposium on –

IIIX 12, 164.

Villa, R., Halvey, M., Joho, H., Hannah, D., & Jose, J. M. (2010). Can an intermediary collection

help users search image databases without annotations? Proceedings of the 10th annual joint

conference on Digital libraries

Westman, S., Lustila, A., & Oittinen, P. (2008). Search strategies in multimodal image retrieval. In

M. Lalmas, A. Tombros, P. Borlund, J. W. Schneider, D. Kelly, J. Feather, & A. P. de Vries (Eds.),

Proceedings of the 2nd International Conference on Information Interaction in Context (pp. 13–

20). New York: NY: Association for Computing Machinery.

Yoon, J. (2011). Searching images in daily life. Library & Information Science Research, 33(4), 269–

275.

Zhang, L., Chen, L., Jing, F., Deng, K. and Ma, W. (2006). EnjoyPhoto—A Vertical Image Search

Engine for Enjoying High-Quality Photos. Proceedings of the 14th annual ACM international

Page 44: People’s preferences, perceptions and frequency of using

44

conference on Multimedia, 367-376.

Page 45: People’s preferences, perceptions and frequency of using

45

Appendix A Questionnaire

Frequency of using search engines to

search for images

*Required

Information Consent

The University of Sheffield. Information School Researchers: Junyu Li MSc

Information System Management [email protected] Doctor Robert Villa

Lecturer of Information School [email protected] We wish to better

understand people’s experience of carrying out image searches, including people’s

preferences, perceptions and frequency of use of image search engines. This

study aims to investigate how people use image search engine, people’s behaviour

using image retrieval systems and why people succeed or fail searching for images.

This study would be important for building image search systems and will help to

improve the image search systems in the future. We are inviting adults who used

search engines to search images (pictures, photos, maps etc) before. If you are

not, please ignore this survey. The risks of participating are no greater than those

experienced in everyday life. We are collecting only the results of the

questionnaires and no personal identification will be collected. No identifying

information will be collected. All responses will be anonymous. The data will be

stored at the University of Sheffield on a secure server. Your participation will be

kept strictly confidential and the information you provide will only be used for

my dissertation which will be publicly available. Please contact the School in six

months. If you require any addition information, please contact one of the

researchers listed above. I understand that my participation is voluntary and that

I am free to withdraw at any time without any negative consequences. I

understand that I may decline to answer any particular question or questions, by

closing the browse window. I understand that my responses will be kept strictly

confidential, that my name or identity will not be linked to any research materials,

and that I will not be identified or identifiable in any report or reports that result

from the research. I give permission for the research team members to have

access to my anonymised responses. I give permission for the research team re-

use my data for further research as specified above. I agree to take part in the

research project as described above. [Note: If you have any difficulties with, or

wish to voice concern about, any aspect of your participation in this study, please

contact Dr. Angela Lin, Research Ethics Coordinator, Information School, The

Page 46: People’s preferences, perceptions and frequency of using

46

University of Sheffield ([email protected]), or to the University

Registrar and Secretary.]

By pressing continue *

I confirm I have read the above information sheet, and I agree to take part in the

research project as described above

o I Confirm

Part One

Basic Information about you

1. Please select your gender? *

o Male

o Female

2. Please select your age group *

o 18-24

o 25-30

o 31-40

o 41-50

o Over 50

3. Are you currently or primarily (more than 50% of time) a/an ______? *

o Employed

o Student

o Unemployed

o Other:

4. Please select your educational background *

o High school / Secondary School

o Further education/ College diploma

o Undergraduate

o Master

o Doctorate

o Professional (law,medicine...)

o Other:

Page 47: People’s preferences, perceptions and frequency of using

47

Part Two

Basic Questions about image retrieves

5. On average, how often do you use search engines to search for images? *

(Images can include pictures, photos, maps, etc.)

o 1 to 5 times per week

o 5 to 10 times per week

o 10 to 20 times per week

o More than 20 times per week

6. How many times per week do you use search engine to search for images in your

job or study? *

o 1 to 5 times per week

o 5 to 10 times per week

o 10 to 20 times per week

o More than 20 times per week

7. How many times per week do you use search engine to search for images

OUTSIDE of your job or study? *

o 1 to 5 times per week

o 5 to 10 times per week

o 10 to 20 times per week

o More than 20 times per week

8. How important is image search to you for the following tasks? *

Very

important Important

Neither important

nor unimportant

Unimportant Very

unimportant

Prepare presentation materials

Use as computer wallpaper

Get ideas for purchasing

Travel destinations

Beauty/Fashion

Page 48: People’s preferences, perceptions and frequency of using

48

Very

important Important

Neither important

nor unimportant

Unimportant Very

unimportant

items

See what something looks like

Verify names

Out of Curiosity

Are there any other tasks for which you conduct image searches?

9. Which search engines do you normally use to search for images? *

(Please click as many responses as apply)

o Google (web search)

o Google (search by image)

o Yahoo!

o Tineye

o Baidu (Chinese)

o Bing

o Yandex (Russian)

o Ask

o AOL

o Other:

10. Which of the following are most important to you when conducing image

searchers? *

Strongly

Agree Somewhat

Agree

Neither Agree Nor Disagree

Somwhat Disagree

Strongly Disagree

Good search

Page 49: People’s preferences, perceptions and frequency of using

49

Strongly

Agree Somewhat

Agree

Neither Agree Nor Disagree

Somwhat Disagree

Strongly Disagree

results

Familiarity with the search engine

User-friendly interface

The number of search results and their variety

Availability of relevant text information

Availability of copyright-unrestricted images

Ability to share images with other people

Part Three

Questions about image retrieves experience

11. When was your most recent search for an image *

o Within an hour

o 1-2 hours ago

o More than 3 hours ago but within a day

o 1-3 days ago

o More than 4 days ago but within a week

Page 50: People’s preferences, perceptions and frequency of using

50

o More than one week ago

12. Why did you carry out this search *

o Curiosity

o Entertainment

o Job/ academic needs

o Gain knowledge

o Problem solves (e.g search for a travel destination, map, etc.)

o Other:

13. What was the topics on which you searched? *

o News related material

o Plants/Animals

o Cook/Food related material

o Fashion/Beauty items

o Arts related material

o Sports related material

o Places (e.g travel destination, map, etc )

o Mechanism related material (e.g motocycle, car, etc)

o Electronics products (e.g laptop, phones, etc)

o Information thechnology related material

o Medical related material

o Films, Books or Music related material

o Figure of People (e.g celebrity, etc)

o Other:

14. What was the search engine you used for this search *

o Google (web search)

o Google (search by image)

o Yahoo!

o Tineye

o Baidu (Chinese)

o Bing

o Yandex (Russian)

o Ask

o AOL

Page 51: People’s preferences, perceptions and frequency of using

51

o Other:

15. How difficult was this search, to find relevant images? *

1 2 3 4 5

Easy Select a value from a range from 1,Easy, to 5,Di ffi cult,. Difficult

16. How happy were you with this search results? *

o Very successful, I found exactly what I required (Jump to Q18)

o Successful (Jump to Q18)

o Neither successful nor failure

o Failure

o Very failure, I did not find what I required

17. Please rate the following statements, if you felt you were not successful in your

image search (Jump to Q19)

Strongly

Agree Somewhat

Agree

Neither Agree Nor Disagree

Somwhat Disagree

Strongly Disagree

Images were not relevant

Images with low resolution

Images with wrong size

Images were unfamiliar that can't be told whether it was required

No clear copyrights of images

Image search options or interface

Page 52: People’s preferences, perceptions and frequency of using

52

Strongly

Agree Somewhat

Agree

Neither Agree Nor Disagree

Somwhat Disagree

Strongly Disagree

were difficult to use

Please enter any other reasons you felt led to an unsuccessful image search

18. Please rate the following statements, if you felt you were successful in your image

search

Strongly

Agree Somewhat

Agree

Neither Agree Nor Disagree

Somwhat Disagree

Strongly Disagree

Images were highly relevant

Results of search came out quickly

Search engine was easy to operate

The search engines represented in your native languange

Images with clear copyrights

Images with high resolution

Please enter any other reasons you felt led to a successful image search

Page 53: People’s preferences, perceptions and frequency of using

53

19. Please select ways the image search engine you used can be improved *

(Please click as many responses as apply)

o Returning more relevant images and fewer irrevelant images

o Improving the quality or size of the returned images

o Providing more search options or improving the interface

o Providing related information about the image

o Improving the way that returned images are displayed

o Displaying copyrighted images or images from commercial websites

o Other:

Appendix B Ethics Proposal Students Staff This proposal submitted by: This proposal is for:

Undergraduate Specific research project

X Postgraduate (Taught) – PGT Generic research project

Postgraduate (Research) – PGR This project is funded by:

N/A

Project Title: People’s preferences, perceptions and frequency of using search engines to search for images

Start Date: July 15, 2013 End Date: August 30, 2013

Principal Investigator (PI):

(student for supervised UG/PGT/PGR research)

Junyu Li

Email: [email protected]

Supervisor:

(if PI is a student)

Doctor Robert Villa

Email: [email protected]

Indicate if the research: (put an X in front of all that apply)

Page 54: People’s preferences, perceptions and frequency of using

54

Involves adults with mental incapacity or mental illness, or those unable to make a personal decision

Involves prisoners or others in custodial care (e.g. young offenders)

Involves children or young people aged under 18 years of age

Involves highly sensitive topics such as ‘race’ or ethnicity; political opinion; religious, spiritual or other beliefs; physical or mental health conditions; sexuality; abuse (child, adult); nudity and the body; criminal activities; political asylum; conflict situations; and personal violence.

Please indicate by inserting an “X” in the left hand box that you are conversant with the University’s policy

on the handling of human participants and their data.

X

We confirm that we have read the current version of the University of Sheffield Ethics Policy

Governing Research Involving Human Participants, Personal Data and Human Tissue, as shown on

the University’s research ethics website at: www.sheffield.ac.uk/ris/other/gov-ethics/ethicspolicy

Page 55: People’s preferences, perceptions and frequency of using

55

Part B. Summary of the Research

B1. Briefly summarise the project’s aims and objectives: (This must be in language comprehensible to a layperson and should take no more than one-half page. Provide enough information so that the reviewer can understand the intent of the research) Summary:

Image search is used in many different contexts for many different purposes: for example, some

image search behaviour may originate from a user’s daily life, from a job task, or from academic life.

In this study, I will investigate people’s preferences, perceptions and frequency of using search

engines to search for images, which can help to understand people’s image search requirements, and

how they search images through search engines. This study will investigate people’s most recent

experience of image search. The objectives are to investigate how people use image search engines,

people’s behaviours using image retrieval systems, and why people succeed or fail when conducting

image searches. This study will be important for building image search systems and will help to

improve the image search systems in the future.

B2. Methodology: Provide a broad overview of the methodology in no more than one-half page.

Overview of Methods:

This study will conduct a web-based survey created on Google Docs, and all undergraduate,

postgraduate, doctoral students and staff in the University of Sheffield would be invited to complete

the survey.

If more than one method, e.g., survey, interview, etc. is used, please respond to the questions in Section C for each method. That is, if you are using both a survey and interviews, duplicate the page and answer the questions for each method; you need not duplicate the information, and may simply indicate, “see previous section.”

Page 56: People’s preferences, perceptions and frequency of using

56

C1. Briefly describe how each method will be applied Method (e.g., survey, interview, observation, experiment): The methods used will be web-based survey only. Description – how will you apply the method? The survey will be created and mounted on Google Docs and available from July 15-30, 2013. The survey will conclude three parts: Part 1 consists of four questions about education, gender, age, etc. to investigate participants’ basic information and draw a demographic profile of them. Part 2 will contain 6 questions about people’s preferences and frequency of using image search engines in both their daily and academic life. Part 3 will contain 8 questions about the participants’ most recent experience of using image search engines including their perceptions and reasons. Most of those questions are closed questions.

About your Participants C2. Who will be potential participants? Potential participants will be undergraduate, postgraduate, doctoral students and staff enrolled in the University of Sheffield C3. How will the potential participants be identified and recruited? The potential participants will be sent an email via the VOLUNTEERS discussion list by Doctor Villa C4. What is the potential for physical and/or psychological harm / distress to participants? The potential for harm or distress is no larger than what might be experienced in everyday life C5. Will informed consent be obtained from the participants?

X Yes

No

If Yes, please explain how informed consent will be obtained? The Information Sheet/Consent Form will be the first page of the survey and will describe what will be asked of participants. An email will be supplied for further information if a participant has questions that are not answered on the Information Sheet. If No, please explain why you need to do this, and how the participants will be de-briefed? C6. Will financial / in kind payments (other than reasonable expenses and compensation for time) be offered to participants? (Indicate how much and on what basis this has been decided) No financial/in kind payments will be provided

About the Data C7. What data will be collected? (Tick all that apply)

Print Digital

Participant observation

Audio recording

Video recording

Computer logs

Questionnaires/Surveys X

Other:

Other:

Page 57: People’s preferences, perceptions and frequency of using

57

C8. What measures will be put in place to ensure confidentiality of personal data, where appropriate? Each survey response will be associated with an arbitrary number, and all responses will be anonymous. C9. How/Where will the data be stored? The data will be exported from Google Docs and safely stored on my laptop and the University of Sheffield secure server which can only accessed by myself and Doctor Villa. C10. Will the data be stored for future re-use? If so, please explain The data will be only used for my postgraduate project, and other academic research purposes.

About the Procedure C11. Does your research raise any issues of personal safety for you or other researchers involved in the project (especially if taking place outside working hours or off University premises)? If so, please explain how it will be managed. There are no issues of personal safety involved with this survey and there will be no contact between the investigators and the participants. The survey will be anonymous.

Page 58: People’s preferences, perceptions and frequency of using

58

The University of Sheffield. Information School

Research Ethics Review Declaration

Title of Research Project: People’s behaviours through image retrieve

We confirm our responsibility to deliver the research project in accordance with the University of

Sheffield’s policies and procedures, which include the University’s ‘Financial Regulations’, ‘Good

Research Practice Standards’ and the ‘Ethics Policy Governing Research Involving Human

Participants, Personal Data and Human Tissue’ (Ethics Policy) and, where externally funded, with the

terms and conditions of the research funder.

In submitting this research ethics application form I am also confirming that:

The form is accurate to the best of our knowledge and belief.

The project will abide by the University’s Ethics Policy.

There is no potential material interest that may, or may appear to, impair the independence

and objectivity of researchers conducting this project.

Subject to the research being approved, we undertake to adhere to the project protocol

without unagreed deviation and to comply with any conditions set out in the letter from the

University ethics reviewers notifying me of this.

We undertake to inform the ethics reviewers of significant changes to the protocol (by

contacting our academic department’s Ethics Coordinator in the first instance).

we are aware of our responsibility to be up to date and comply with the requirements of

the law and relevant guidelines relating to security and confidentiality of personal data,

including the need to register when necessary with the appropriate Data Protection Officer

(within the University the Data Protection Officer is based in CiCS).

We understand that the project, including research records and data, may be subject to

inspection for audit purposes, if required in future.

We understand that personal data about us as researchers in this form will be held by those

involved in the ethics review procedure (e.g. the Ethics Administrator and/or ethics

reviewers) and that this will be managed according to Data Protection Act principles.

If this is an application for a ‘generic’ project all the individual projects that fit under the

generic project are compatible with this application.

We understand that this project cannot be submitted for ethics approval in more than

one department, and that if I wish to appeal against the decision made, this must be done

through the original department.

Name of the Student (if applicable):

Junyu Li

Name of Principal Investigator (or the Supervisor):

Doctor Robert Villa

Date: July7, 2013

Page 59: People’s preferences, perceptions and frequency of using

59

Appendix C Ethics Information Consent The University of Sheffield. Information School

People’s preferences, perceptions and frequency of using search engines to search for images

Researchers

Junyu Li Master of Informatics Student Information School University of Sheffield [email protected]

Doctor Robert Villa Lecturer of Information School Information School University of Sheffield [email protected]

Purpose of the research

We wish to better understand people’s experience of carrying out image searches, including people’s preferences, perceptions and frequency of use of image search engines. This study aims to investigate how people use image search engine, people’s behaviour using image retrieval systems and why people succeed or fail searching for images. This study would be important for building image search systems and will help to improve the image search systems in the future.

Who will be participating?

We are inviting undergraduate, postgraduate, doctoral students and staff enrolled in the University of Sheffield to participate in a web-based survey.

What will you be asked to do?

We will ask you to respond to a series of questions in three parts: Part 1 consists of four questions about education, gender, age, etc. to investigate participants’ basic information and draw a demographic profile. Part 2 will contain 6 questions about people’s preferences and frequency of using image search engines in both their daily and academic or work life. Part 3 will contain 9 questions about participants’ most recent experience of using image search engines including their perceptions and reasons. The survey will take about 10 minutes to complete.

What are the potential risks of participating?

The risks of participating are the same as those experienced in everyday life

What data will we collect?

We are only collecting your responses to the survey questions. No other data will be recorded.

What will we do with the data?

We will be analysing the data for inclusion in my master’s dissertation and other research purposes.

Will my participation be confidential?

The data is anonymous and all responses will be associated with a number. No personal identifying information will be recorded or retained.

What will happen to the results of the research project?

The results of this study will be included in my master’s dissertation which will be publicly available. Please contact the School in six months.

Page 60: People’s preferences, perceptions and frequency of using

60

I confirm that I have read and understand the description of the research project, and that I have had an opportunity to ask questions about the project. I understand that my participation is voluntary and that I am free to withdraw at any time without any negative consequences. I understand that I may decline to answer any particular question or questions, or to do any of the activities. If I stop participating at all time, all of my data will be purged. I understand that my responses will be kept strictly confidential, that my name or identity will not be linked to any research materials, and that I will not be identified or identifiable in any report or reports that result from the research. I give permission for the research team members to have access to my anonymised responses. I give permission for the research team to re-use my data for future research as specified above. I agree to take part in the research project as described above.

Participant Name (Please print) Participant Signature

Researcher Name (Please print) Researcher Signature Date

Note: If you have any difficulties with, or wish to voice concern about, any aspect of your participation in this study, please contact Dr. Angela Lin, Research Ethics Coordinator, Information School, The University of Sheffield ([email protected]), or to the University Registrar and Secretary.

Page 61: People’s preferences, perceptions and frequency of using

61

Appendix D Approval Letter

Information School Research Ethics Panel Letter of Approval Date: 12th July 2013 TO: Junyu Li The Information School Research Ethics Panel has examined the following application: Title: People’s preferences, perceptions and frequency of using search engines to search for images

Submitted by: Junyu Li And found the proposed research involving human participants to be in accordance with the University of Sheffield’s policies and procedures, which include the University’s ‘Financial Regulations’, ‘Good Research Practice Standards’ and the ‘Ethics Policy Governing Research Involving Human Participants, Personal Data and Human Tissue’ (Ethics Policy). This letter is the official record of ethics approval by the School, and should accompany any formal requests for evidence of research ethics approval. Effective Date: 12th July 2013

Dr Angela Lin Research Ethics Coordinator

Page 62: People’s preferences, perceptions and frequency of using

62

Access to Dissertation A Dissertation submitted to the University may be held by the Department (or School) within which the Dissertation was undertaken and made available for borrowing or consultation in accordance with University Regulations. Requests for the loan of dissertations may be received from libraries in the UK and overseas. The Department may also receive requests from other organisations, as well as individuals. The conservation of the original dissertation is better assured if the Department and/or Library can fulfill such requests by sending a copy. The Department may also make your dissertation available via its web pages. In certain cases where confidentiality of information is concerned, if either the author or the supervisor so requests, the Department will withhold the dissertation from loan or consultation for the period specified below. Where no such restriction is in force, the Department may also deposit the Dissertation in the University of Sheffield Library.

To be completed by the Author – Select (a) or (b) by placing a tick in the appropriate box If you are willing to give permission for the Information School to make your dissertation available in these ways, please complete the following:

√ (a) Subject to the General Regulation on Intellectual Property, I, the author, agree to this dissertation being made immediately available through the Department and/or University Library for consultation, and for the Department and/or Library to reproduce this dissertation in whole or part in order to supply single copies for the purpose of research or private study

(b) Subject to the General Regulation on Intellectual Property, I, the author, request that this dissertation be withheld from loan, consultation or reproduction for a period of [ ] years from the date of its submission. Subsequent to this period, I agree to this dissertation being made available through the Department and/or University Library for consultation, and for the Department and/or Library to reproduce this dissertation in whole or part in order to supply single copies for the purpose of research or private study

Name Junyu Li

Department Information School

Signed Date 25/08/2013

To be completed by the Supervisor – Select (a) or (b) by placing a tick in the appropriate box

(a) I, the supervisor, agree to this dissertation being made immediately available through the Department and/or University Library for loan or consultation, subject to any special restrictions (*) agreed with external organisations as part of a collaborative project.

*Special restrictions

(b) I, the supervisor, request that this dissertation be withheld from loan, consultation or reproduction for a period of [ ] years from the date of its submission. Subsequent to this period, I, agree to this dissertation being made available through the Department and/or University Library for loan or consultation, subject to any special restrictions (*) agreed with external organisations as part of a collaborative project

Name

Department

Signed Date

THIS SHEET MUST BE SUBMITTED WITH DISSERTATIONS IN ACCORDANCE WITH DEPARTMENTAL REQUIREMENTS.