WIMS’ 14

Providing a context-aware location based web service

through semantics and user-defined rules

Iosif Viktoratos1, Athanasios Tsadiras1, Nick Bassiliades2,

1Department of Economics, 2Department of Informatics, Aristotle University of ThessalonikiGR-54124 Thessaloniki, Greece

{viktorat, tsadiras, nbassili}@auth.gr

Contents

Location Based Services (LBS) & Context Semantic Technologies The System Geo SPLIS

Design and General idea Geo SPLIS’s Features Geos SPLIS’s Architecture Geo SPLIS’s Operations

Usage Scenarios Evaluation Conclusions Future work

2/36

Location Based Services (LBS)

Very popular over the last few years

Used by millions of people Navigation

Tracking

Information

Marketing

Entertainment

Emergency

Communication

3/36

LBS & Context

Should offer information to users, relevant to their situation - context Time, Location, User Preferences, Relationships…

Context awareness enables proactive personalized information of higher quality

Researchers and industries focus on contextual knowledge Hardware (e.g. GPS, sensors)

Software (e.g. ontologies,rules)

4/36

5/36

Semantic Technologies (1/2)Ontologies

Representation of related concepts (persons’ profiles, location) High quality context perception

Complicated reasoning about concepts

Formal representation standard Reusability

Interoperability

Connecting and sharing data from various sources Flexibility

Knowledge sharing

6/36

Semantic Technologies (2/2)Rules

Extensive reasoning capabilities Reasoning about instances

Intelligence Increased expressiveness for complicated

inferences

Autonomy - Proactiveness Data-driven inference

Automatic derivation upon context changes

7/36

Geo SPLISGeographic Semantic Personalized Location Information System

What? Web mapping application for connecting user defined

preferences (regarding POIs) with POI owners’ group targeted offers

Why? To model human daily patterns and provide

proactive, customized and contextualized information to users

How? Combining semantics (rules & ontologies) with LBS

8/36

Design and General idea (1/2)

POIs adopt a rule-based policy to deploy their specific marketing strategy A restaurant offers 25% discount to students on Friday

Regular users provide their profile My job is student

Regular users have preferences/daily patterns If it is Sunday noon I would like some restaurants that

serve Chinese cuisine The LBS provides spatiotemporal context

E.g. its Friday! Combine POI policies with regular user preferences

& spatiotemporal context Presents personalized offers on Google Maps

9/36

Design and General idea (2/2)

Geo SPLIS

Rule-based group targeted offers and

POI data

POI owner

Rule-based preferences and

Contextual data

Contextualized information

Regular user

10/36

Geo SPLIS’s Features (1/2) Collects data from external sources

Google+, Google Places API, POI websites Regular users add contextualized rule based

preferences Via a web editor Upload a RuleML file in their Google+ account

POI owners add group targeted offering policies via a web editor

Data from editor RuleML Jess Sesame Executes and evaluates data and rules on the fly Uses Google Maps for information

Geo SPLIS’s Features (2/2) Rule conditions

LBS context Location (e.g. within 1 km) Weather (e.g. sunny, rainy etc. ) Time (e.g. between 14:00-18:00) Day (e.g. Saturday)

Every existing property of a POI E.g. cuisine currently serves

Rule consequences Add a place in a recommendation list

11/36

Geo SPLIS’s Architecture (1/2)

Client PC browser-based

Html, JavaScript, Css Google Maps

Android mobile application

Server Java Server Pages (JSP)

RDF data management Sesame

Rules Reaction RuleML XSLT Jess

12/36

13/36

Geo SPLIS’s Architecture (2/2)

GeoSPLIS Server

External sources(Scema.org, POI websites, Google API)

SesameJess

GeoSPLIS Client

Geo SPLIS’s Operations(1/11)

Data collection

Information presentation

Operations concerning rules Rule insertion through editor

Rule insertion through Google+

Rule update operation

“Get a rule operation”

Planning operation

14/36

15/36

Geo SPLIS’s Operations(2/11)Data collection

Ontology loading Load the schema.org ontology into Sesame for updates

Data update Get user data either from a registration form or from

Google+ account

Obtain user position and retrieve nearest POIs from Google Places API

Store data for POIs into Sesame using predefined mapping to schema.org

16/36

Geo SPLIS’s Operations (3/11)Information presentation

Data retrieval Fetch user’s profile data and rules (if any) Calculate contextual property values (location etc.) Fetch POIs’ property values and rules (if any)

Rule evaluation Assert above data to the Jess rule engine Evaluate rules using the asserted facts

Presentation of personalized information Bigger in size marker recommended POI Green marker at least one valid offer for the user Yellow marker no rule is fired for the current user Red marker no offers at all Red star POI owner

17/36

Mobile version

Geo SPLIS’s operations (4/11)Information presentation layer

PC browser-based version

Geo SPLIS’s Operations (5/11) Rule insertion through editor

Enter the title and the priority of the rule Four “Add …. Condition” buttons

Each one for the corresponding contextual condition. The condition customization consists of three elements:

The property field (weather, day, time, distance) The operator field (“is” and “<”,”>” for time and distance) The value An “AND” is implied among them Drop down menus and read only forms to avoid mistakes

Select the desired POI category Clicking the “Add Where Condition” button to add more

conditions regarding desired POI properties Add a textual explanation Assert the rule “If day is Friday and time is before

17:00, I would like to visit some museums” 18/36

Geo SPLIS’s operations (6/11) Geo SPLIS’s Web Editor

(defrule museums (declare (salience 1))(place( type Museum) ( uri ?id)) (person ( day friday) ( time ?t) ) (test (< ?t 17))=> (assert (recommendation( id ?id))) (store EXPLANATION " If day is Friday and time is before 17:00, I would like to visit some museums "))

Jess representation

RuleML representation

20/36

Geo SPLIS’s Operations (8/11)Rdf data representation

Geo splis: policy19d883ef-f735-4521-a8a6 771165b1b2a8

rdf:type Schema:Policy;

schema: policy_description

IF person:time < 17 AND person:day is Friday THEN I WOULD LIKE TO GO TO A place:type Museum

schema:policy_explanation

If day is Friday and time is before 17:00, I would like to visit some museums

schema:policy_priority

1

schema:policy_link

platon.econ.auth.gr/......ruleml

schema: Person11 schema:Policy Geo splis: policy19d883ef-f735-4521-a8a6 771165b1b2a8

21/36

Geo SPLIS’s Operations (9/11)Rule insertion through Google+

1. Set as label the text “policy_link”

2. Upload a ruleml link in Google+

3. Geo SPIS crawls user page, parses the rules and evaluates them on the fly

22/36

Geo SPLIS’s Operations (10/11)Operations concerning rules

Rule update operation Edit and delete a rule through editor

“Get a rule” operation Get rules from other users

Check and get the rule

In case of editing a rule, a user is simply “unlinked” from the rule so that not to affect other users

23/36

Geo SPLIS’s Operations (11/11)Planning operation

A user inserts a future day (1-5 days)

The time of day

The system gets user’s future contextual situation

Geo SPLIS provides POIs and offers regarding the future situation

Usage scenario will presented

24/36

Usage Scenarios (1/6)Normal usage

A scenario concerning a selected user (“Mary”) Personalized information Planning operation

Rule Explanation

Rule 1

If day is Wednesday and time is between 12:00-17:00, then I would like some restaurants

Rule 2

If day is Friday and time is before 17:00, I would like to visit some museums

Name Age JobTitle Time Day Weather Location

Mary 20 Student 14:25 Wednesday Sunny Location A

Usage Scenarios (2/6) Normal usage

Because its Wednesday, 14:25 Mary’s rule 1 is fired

As a result restaurants are represented with a bigger in size marker

27/36

Usage Scenarios (3/6) Normal usage

POI “Nama” is represented with a bigger marker because it is a restaurant (rule 1)

Mary is entitled for a special spaghetti price as a student (POI owner’s rule)

28/36

Usage Scenarios (4/6) Normal usage

POI “Pizza Greca” is represented with a bigger marker because it is a restaurant

Mary is not entitled for a special pizza price as a student (POI owner’s rule)

Usage Scenarios (5/6) Planning mode

Mary will visit Rome on Friday afternoon

Clicks “Planning” button from menu

Selects the following options in the pop window after 2 days (it’s Wednesday)

the time she will be there (13:00 in our scenario)

As soon as she is inserted into planning mode she can right click on the map in Rome and find places and offers matching the future situation

29/36

Usage Scenarios (6/6) Planning mode

The rule “If day is Friday and time is before 17:00, I would like to visit some museums” is fired (It’s Friday 13:00 o’clock)

Museums are represented with a bigger in size markerOffers regarding the future situation

Evaluation(1/3)

Quantitative evaluation was made Geo SPLIS compared with a stripped-

down version Only yellow and red markers Users had to click on the marker, view

POI’s category to understand if it matches his/her preference and read POI owner’s message to understand if their offers matches his/her profile or not

“If day is Wednesday, find me some Coffee shops”

In stripped-down it was just told to users

31/36

Evaluation (2/3)

44 undergraduate students of Economics (18-22 years old, both genders)

3 tasks in both systems Task1.Find how many places match with your preference and

contain an offer which also matches with your profile. Task2.Find how many places contain an offer which matches

with your profile. Task3.Find how many places match with your preference and

contain an offer that does not match with your profile. Equal number of solutions

25 places 2 places/solutions to Task 1 6 places/solutions to Task 2 3 places/solutions to Task 3

32/36

Evaluation (3/3)

0

5

10

15

20

25

30

35

Task 1 Task 2 Task 3

Median time for the stripped-down version of Geo SPLIS

Average time for the strippeddown version of Geo SPLIS

Median time for Geo SPLIS

Average time for Geo SPLIS

Significant difference (p-significant was calculated 0,00131<0,05) 33/36

Conclusions

Geo SPLIS models human daily preferences and connect them with POI owners group based offering policies

Advantages POI owners have highly targeted marketing audience Regular users enjoy proactive contextualized information Adding rules dynamically leads to more customized and

personalized user experience The more rules were added to the system, the more

interesting/intelligent it becomes Exploit people collaboration and social intelligence to

create large knowledge bases34/36

Future work

More extensive user evaluation Add a social layer

Combine user rules with those of their logged in friends and provide group based information

Manually using friends’ rules When personal preferences do not fire, then

automatically use social graph to recommend Expand the editor to add a wider range of

preferences (movies, music etc.) Rules (semi-)automatically induced by mining

users’ logs, likes or reviews Connect other social media sources (facebook,

twitter etc.) 35/36

Thanks for your time!

Geo SPLIS is available at: http://tinyurl.com/GeoSPLIS

Mobile version is available at: http://tinyurl.com/GeoSoSPLIS36/36