interactive recommender systems

107
Interactive Recommender Systems: Bridging the gap between predictive algorithms and interactive user interfaces Denis Parra, Ph.D. Information Sciences Assistant Professor, CS Department School of Engineering Pontificia Universidad Católica de Chile UFMG, March 29 th 2017

Upload: denis-parra-santander

Post on 11-Apr-2017

180 views

Category:

Education


2 download

TRANSCRIPT

Page 1: Interactive Recommender Systems

Interactive Recommender Systems:Bridging the gap between predictive

algorithms and interactive user interfaces

Denis Parra, Ph.D. Information SciencesAssistant Professor, CS DepartmentSchool of EngineeringPontificia Universidad Católica de Chile

UFMG, March 29th 2017

Page 2: Interactive Recommender Systems

Outline

• Brief Personal Introduction• Computer Science at PUC Chile• Projects at SocVis Lab

• Overview of Recommender Systems • Interactive Recommender Systems

• Summary & Current & Future Work

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 2

Page 3: Interactive Recommender Systems

1 slide Geography Class: Chile• One third of the 16 million Chileans lives in Santiago,

the Capital• But Chile is a looong country (4.000 Km), in the north

is hot and dry, in the south (Patagonia) is very cold.Very Hot!

Very Cold!

My hometown!Valdivia

Santiago, PUC Chile

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 3

Page 4: Interactive Recommender Systems

Personal Introduction 1/3

• B.Eng. and Engineering in Informatics from Universidad Austral de Chile (2004), Valdivia, Chile

• Ph.D. in Information Sciences at University of Pittsburgh (2008-2013), Pittsburgh, PA, USA

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 4

Page 5: Interactive Recommender Systems

Personal Introduction 2/3

• In 2009 I did an internship at Trinity College Dublin, with researcher Alexander Troussov (IBM)

• In 2010 I did another internship at Telefonica I+D, Barcelona, with Xavier Amatrian (now VP Quora)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 5

Page 6: Interactive Recommender Systems

Personal Introduction 3/3

• 2013: Moved back to Santiago, Chile • Department of CS, School of Engineering, PUC.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 6

Page 7: Interactive Recommender Systems

DCC, Engineering, PUC Chile• DCC: Departamento de Ciencia de la Computación• Programs: BEng, Engineering title, Master, PhD• Research Areas:

– Databases and Semantic Web– Information Technologies– Machine Learning and Computer Vision (GRIMA)– Software Engineering– Educational Technologies, MOOCs

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 7

http://dcc.ing.puc.cl

Page 8: Interactive Recommender Systems

Academic activities (2017)

• Research topics: Recommender Systems/Personalization, Visualization, SNA.

• Teaching: Data Mining, Recommender Systems, Information Visualization, SNA.

• Leading the Social Computing and Visualization (SocVis) Lab.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 8

Page 9: Interactive Recommender Systems

SocVis Lab

http://www.socvis.clMarch 29th, 2017 D.Parra ~ UFMG – Invited Talk 9

Page 10: Interactive Recommender Systems

People, Publications, News (ND)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 10

Page 11: Interactive Recommender Systems

Projects at SocVis

• Mood-based music artists recommendation– Collaboration with J. O’Donovan (UCSB)– Student: Raimundo Herrera

• IR on evidence-based Medicine– Help doctors on answering clinical questions– Student: I. Donoso, collaboration Epistemonikos

• Artwork Recommendation – Collaboration with online artwork store UGallery– Students: P. Messina & V. Dominguez

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 11

Page 12: Interactive Recommender Systems

Recommender Systems Class

• Recommender Systems at PUC Chilehttp://web.ing.puc.cl/~dparra/classes/recsys-2016-2/

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 12

Page 13: Interactive Recommender Systems

INTRODUCTION TO RECSYSRecommender Systems

* Danboard (Danbo): Amazon’s cardboard robot, in these slides it represents a recommender system

*

Page 14: Interactive Recommender Systems

Recommender Systems (RecSys)Systems that help (groups of) people to find relevant items in

a crowded item or information space (MacNee et al. 2006)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 14

Page 15: Interactive Recommender Systems

Why do we care about RecSys?

• RecSys have gained popularity due to several domains & applications that require people to make decisions among a large set of items.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 15

Page 16: Interactive Recommender Systems

A lil’ bit of History

• First recommender systems were built at the beginning of 90’s (Tapestry, GroupLens, Ringo)

• Online contests, such as the Netflix prize, grew the attention on recommender systems beyond Computer Science(2006-2009)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 16

Page 17: Interactive Recommender Systems

The Recommendation Problem

• The most popular way of presenting the recommendation problem is rating prediction:

• How good is my prediction?

Item 1 Item 2 … Item m

User 1 1 5 4

User 2 5 1 ?

User n 2 5 ?

Predict!

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 17

Page 18: Interactive Recommender Systems

Recommendation Methods

• Without covering all possible methods, the two most typical classifications on recommender algorithms are

Classification 1 Classification 2- Collaborative Filtering- Content-based Filtering- Hybrid

- Memory-based- Model-based

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 18

Page 19: Interactive Recommender Systems

Collaborative Filtering (User-based KNN)• Step 1: Finding Similar Users (Pearson Corr.)

5

4

4

1

2

1

54

4

12

5

Active user

User_1

User_2

User_3

active user

user_1

user_2

user_3

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 19

Page 20: Interactive Recommender Systems

Collaborative Filtering (User-based KNN)• Step 1: Finding Similar Users (Pearson Corr.)

5

4

4

1

2

1

54

4

12

5

Active user

User_1

User_2

User_3

∑∑∑

⊂⊂

−−

−−=

nunu

nu

CRi nniCRi uui

CRi nniuui

rrrr

rrrrnuSim

,,

,

22 )()(

))((),(

active user

user_1 0.4472136

user_2 0.49236596

user_3 -0.91520863

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 20

Page 21: Interactive Recommender Systems

Collaborative Filtering (User-based KNN)

• Step 2: Ranking the items to recommend

5

4

4

2

1

54

4

Active user

User_1

User_2

2

3

4

2

Item 1

Item 2

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 21

Page 22: Interactive Recommender Systems

Collaborative Filtering (User-based KNN)

• Step 2: Ranking the items to recommend

5

4

4

2

1

54

4

Active user

User_1

User_2

∑∑

⊂−⋅

+=)(

)(

),(

)(),(),(

uneighborsn

uneighborsn nni

u nuuserSim

rrnuuserSimriupred2

3

4

2

Item 1

Item 2

Item 3

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 22

Page 23: Interactive Recommender Systems

Pros/Cons of CFPROS:• Very simple to implement• Content-agnostic• Compared to other techniques such as content-

based, is more accurate. There is also the Item KNN.CONS:• Sparsity• Cold-start• New Item

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 23

Page 24: Interactive Recommender Systems

Content-Based Filtering• Can be traced back to techniques from IR, where

the User Profile represents a query.

user_profile = {w_1, w_2, …., w_3} using TF-IDF, weighting

Doc_1 = {w_1, w_2, …., w_3}

Doc_2 = {w_1, w_2, …., w_3}

Doc_3 = {w_1, w_2, …., w_3}

Doc_n = {w_1, w_2, …., w_3}

5

4

5

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 24

Page 25: Interactive Recommender Systems

PROS/CONS of Content-Based FilteringPROS:• New items can be matched without previous

feedback• It can exploit also techniques such as LSA or LDA• It can use semantic data (ConceptNet, WordNet,

etc.)CONS:• Less accurate than collaborative filtering• Tends to overspecialization

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 25

Page 26: Interactive Recommender Systems

Hybridization• Combine previous methods to overcome their

weaknesses (Burke, 2002)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 26

Page 27: Interactive Recommender Systems

C2. Model/Memory Classification

• Memory-based methods use the whole dataset in training and prediction. User and Item-based CF are examples.

• Model-based methods build a model during training and only use this model during prediction. This makes prediction performance way faster and scalable

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 27

Page 28: Interactive Recommender Systems

Model-based: Matrix Factorization

Latent vector of the itemLatent vector of the user

SVD ~Singular Value Decomposition

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 28

Page 29: Interactive Recommender Systems

PROS/CONS of MF and latent factors model

PROS:• So far, state-of-the-art in terms of accuracy (these

methods won the Netflix Prize)• Performance-wise, the best option nowadays: slow

at training time O((m+n)3) compared to correlation O(m2n), but linear at prediction time O(m+n)

CONS:• Recommendations are obscure: How to explain that

certain “latent factors” produced the recommendation?

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 29

Page 30: Interactive Recommender Systems

Other paradigms and techniques

• Recommendation as a graph problem: – Model the problem as diffusion or link prediction– Personalized PageRank (Kamvar et al, 2010), (Santos et

al 2016)• Recommendation as a ranking problem:

– Rather than predicting ratings, predict a Top-N list– Learning-to-rank approaches developed in the IR

community– Karatzoglou et al. (2013), Shi et al. (2014), Macedo et al.

(2015)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 30

Page 31: Interactive Recommender Systems

(Important) RecSys Topics Not Covered in this Presentation

• Learning to rank• Graph-based methods• Context-aware recommenders• Recommendation problem as next-item in sequence• User-centric evaluation frameworks• Multiarmed Bandits• Reinforcement Learning• ... You need to take Professor Santos’ course J

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 31

Page 32: Interactive Recommender Systems

Rethinking the Recommendation Problem

• User feedback is scarce: need for exploiting different sources of user preference and context

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 32

Page 33: Interactive Recommender Systems

Rethinking the Recommendation Problem

• Ratings are scarce: need for exploiting other sources of user preference

• User-centric recommendation takes the problem beyond ratings and ranked lists: evaluate user engagement and satisfaction, not only RMSE/MAP

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 33

Page 34: Interactive Recommender Systems

Rethinking the Recommendation Problem

• Ratings are scarce: need for exploiting other sources of user preference

• User-centric recommendation takes the problem beyond ratings and ranked lists: evaluate user engagement and satisfaction, not only RMSE/MAP

• Several other dimensions to consider in the evaluation: novelty of the results, diversity, coverage (user and catalog), trust

• Study de effect of interface characteristics: controllability, transparency, explainability.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 34

Page 35: Interactive Recommender Systems

My Take on RecSys Research (2009 ~)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 35

Page 36: Interactive Recommender Systems

My Work on RecSys

• In my research I have contributed to RecSys by:– Utilizing other sources of user preference (Social Tags)– Exploiting implicit feedback for recommendation and for

mapping explicit feedback– Studying interactive interfaces: the effect of visualizations and

user interaction on user satisfaction, perception of trust and accuracy.

• Nowadays: Focus on interactive exploratory interfaces for recommender systems

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 36

Page 37: Interactive Recommender Systems

This is not only My work J• Dr. Peter Brusilovsky

University of Pittsburgh, PA, USA

• Dr. Alexander TroussovIBM Dublin and TCD, Ireland

• Dr. Xavier AmatriainTID / Netflix /Quora

• Dr. Christoff TrattnerNTNU, Norway

• Dr. Katrien VerbertKU Leuven, Belgium

• Dr. Leandro Balby-MarinhoUFCG, Brasil

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 37

Page 38: Interactive Recommender Systems

VISUALIZATION + USER CONTROLLABILITY

Part of this work with Katrien Verbert

Page 39: Interactive Recommender Systems

Human Factors in RecSys• Transparency and Explainability: Konstan et al (2000),

Tintarev and Mastoff (2010)

• Frameworks to evaluate RecSys user studies: ResQue (Pu et al , 2010), Knijnenburg et al (2012)

• Controllability and Inspectability: O’Donovan (2008), Knijnenburg et al (2010, 2012), Hijikata (2012), Ekstrand et al (2015)

• Visualization and Interfaces: O’Donovan (2008 - ..), Verbert et al (2013), Parra et al (2014), Loepp et al (2014, 2017),

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 39

Page 40: Interactive Recommender Systems

Visualization & User Controllability

• Motivation: Can user controllability and explainability improve user engagement and satisfaction with a recommender system?

• Specific research question: How intersections of contexts of relevance (of recommendation algorithms) might be better represented for user experience with the recommender?

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 40

Page 41: Interactive Recommender Systems

Traditional RecSys InterfaceMovieLens: example of traditional recommender list

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 41

Page 42: Interactive Recommender Systems

Explanations and Control

Options of User Control

Explainability

Recommendations of books

GoodReads: Book recommender system

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 42

Page 43: Interactive Recommender Systems

PeerChooser (2008) Controllability in CF

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 43

O’Donovan et al. “PeerChooser: Visual Interactive Recommendation” (2008)

Page 44: Interactive Recommender Systems

SmallWorlds: Expanded Explainability

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 44

Gretarsson et al. “SmallWorlds: Visualizing social recommendations” (2010)

Page 45: Interactive Recommender Systems

TasteWeights: Hybrid Control and Inspect

Bostandjev et al. “TasteWeights: A Visual Interactive Hybrid Recommender System” (2012)

Controllability: Sliders that let users control the importance of preferences and contexts

Inspectability: lines that connect recommended items with contexts and user preferences

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 45

Page 46: Interactive Recommender Systems

IUI 2017

• Loepp et al. (2017)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 46

Page 47: Interactive Recommender Systems

More Details? Check our survey

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 47

He, C., Parra, D., & Verbert, K. (2016). Interactive recommender systems: a survey of the state of the art and future research challenges and opportunities. Expert Systems with Applications, 56, 9-27.

Page 48: Interactive Recommender Systems

Visualization & User Controllability• Motivation: Can user controllability and

explainability improve user engagement and satisfaction with a recommender system?

• Specific research question: How overlapping contexts of relevance (of recommendation algorithms) might be better represented for user experience with the recommender?

• Our scenario: Conference articles

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 48

Page 49: Interactive Recommender Systems

Research Platform

• The studies were conducted using Conference Navigator, a Conference Support System

• Our goal was recommending conference talks

Program Proceedings Author List Recommendations

http://halley.exp.sis.pitt.edu/cn3/

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 49

Page 50: Interactive Recommender Systems

TalkExplorer – IUI 2013• Adaptation of Aduna Visualization to CN• Main research question: Does fusion (intersection) of

contexts of relevance improve user experience?

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 50

Page 51: Interactive Recommender Systems

TalkExplorer - I

EntitiesTags, Recommender Agents, Users

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 51

Page 52: Interactive Recommender Systems

TalkExplorer - II

RecommenderRecommender

Cluster with intersection of entities Cluster (of

talks) associated to only one entity

• Canvas Area: Intersections of Different Entities

User

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 52

Page 53: Interactive Recommender Systems

TalkExplorer - III

ItemsTalks explored by the user

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 53

Page 54: Interactive Recommender Systems

Our Assumptions

• Items which are relevant in more that one aspect could be more valuable to the users

• Displaying multiple aspects of relevance visually is important for the users in the process of item’s exploration

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 54

Page 55: Interactive Recommender Systems

TalkExplorer Studies I & II• Study I

– Controlled Experiment: Users were asked to discover relevant talks by exploring the three types of entities: tags, recommender agents and users.

– Conducted at Hypertext and UMAP 2012 (21 users)– Subjects familiar with Visualizations and Recsys

• Study II– Field Study: Users were left free to explore the interface.– Conducted at LAK 2012 and ECTEL 2013 (18 users) – Subjects familiar with visualizations, but not much with

RecSys

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 55

Page 56: Interactive Recommender Systems

Evaluation: Intersections & Effectiveness

• What do we call an “Intersection”?

• We used #explorations on intersections and their effectiveness, defined as:

Effectiveness =

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 56

Page 57: Interactive Recommender Systems

Results of Studies I & II

• Effectiveness increases with intersections of more entities

• Effectiveness wasn’t affected in the field study (study 2)

• … but exploration distribution was affected

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 57

Page 58: Interactive Recommender Systems

More Details About TalkExplorer• Verbert, K., Parra, D., Brusilovsky, P., & Duval, E. (2013).

Visualizing recommendations to support exploration, transparency and controllability. In Proceedings of the 2013 international conference on Intelligent user interfaces (pp. 351-362). ACM.

• Verbert, K., Parra, D., & Brusilovsky, P. (2016). Agents Vs. Users: Visual Recommendation of Research Talks with Multiple Dimension of Relevance. ACM Transactions on Interactive Intelligent Systems (TiiS), 6(2), 11.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 58

Page 59: Interactive Recommender Systems

SETFUSION: VENN DIAGRAM FOR USER-CONTROLLABLE INTERFACE

Page 60: Interactive Recommender Systems

SetFusion – IUI 2014

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 60

Page 61: Interactive Recommender Systems

SetFusion I

Traditional Ranked List

Papers sorted by Relevance. It combines 3 recommendation approaches.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 61

Page 62: Interactive Recommender Systems

SetFusion - IISlidersAllow the user to control the importance of each data source or recommendation method

Interactive Venn DiagramAllows the user to inspect and to filter papers recommended. Actions available:- Filter item list by clicking on an area- Highlight a paper by mouse-over on a circle- Scroll to paper by clicking on a circle- Indicate bookmarked papers

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 62

Page 63: Interactive Recommender Systems

Study : iConference• A laboratory within-subjects study. 40 subjects.• In Preference elicitation phase, people did not have limit of

papers. Under RecSys interfaces, minimum limit was 15.• In bookmarking, subjects could pick items relevant to a)

themselves, b) themselves and others, and c) only to others.

$12/hourAvg: 1.5 hours

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 63

Page 64: Interactive Recommender Systems

Study : Population and General Stats

Non-controllable Controllable# Total bookmarks 638 625# Average bookmarks/user 15.95 15.63# Average rating 2.48±0.089 2.46±0.076

Gender Female: 17 Male: 23

Age 31.75±6.5Native Speaker Yes: 10 No: 30

Subject Occupation Information Sc. (16), Library Sc.(9), Comp. Sc. (6), Telecomm (3), (+6)

PCA15 questions

on pre-questionnaire

4 Factors (User Characteristics)• Expertise in domain• Engaged with iSchools• Trusting Propensity• Experience w/RecSysDropped• Experience w/CN

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 64

Page 65: Interactive Recommender Systems

Study 2: Results (1)Variables Comment

User Engagement

Significant Talks explored, clicks (nbr. actions) , time spent on task All significantly higher in controllable interface

User Experience

Significant MAP Significantly higher in controllable interface

User Characteristics

Significant Trusting prop.: increases use of Venn diagram and MAPNative speaker: Decreases time spent on taskGender: Being male increases use of sliders Age: Each additional year decreases use of sliders

Trusting propensityconfirms results of previous studies

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 65

Page 66: Interactive Recommender Systems

Rating per method – Effect of Visuals

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 66

Page 67: Interactive Recommender Systems

Gender Differences on SetFusion?

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 67

Page 68: Interactive Recommender Systems

Study : Results (2)Post-session surveys Controllable No-Controllable

Understandability 4.05±0.09*** 2.95±0.16

Satisfaction with interface 4.28±0.09*** 3.4±0.16

Confidence of not missing relevant talks 3.9±0.11*** 3.13±0.15

Intention: I would use it again 4.23±0.09*** 3.45±0.15

Intention: I would recommend system to colleagues 4.28±0.09*** 3.48±0.16

Venn diagram visualization was useful to identify talksrecommended by a specific or by a combination ofrecommendation methods.

4.35±0.11 --

Venn diagram visualization supported explainability4.08±0.13 --

Satisfaction due to ability to control 4.05±0.12 --

Perception of Control with Sliders 4.03±0.13 --

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 68

Page 69: Interactive Recommender Systems

Study : Results (3)Non-control Controllable Both None

Which interface did you prefer? 0 36 4 0Non-control Controllable None Both

Which interface would you suggest to implement permanently in CN?

1 33 1 5

“I like the Venn diagram especially because most papers I was interested in fell in the same intersections, so it was pretty easy to find and bookmark”

“I thought the controllable one adds unnecessary complication if the list is not very long”

“I prefer the sliders (over Venn diagram) because I have used a system before to control search results with a similar widget, so it was more familiar to me.”

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 69

Page 70: Interactive Recommender Systems

Study Takeaways• User Engagement: Controllable interface significantly

drives more user engagement (objective and subjective metrics)

• User Experience: Controllable interface improves user experience by allowing user to interactively control ranking (MAP) and improving explainability.

• User characteristics: Trusting propensity affects positively engagement and experience, engagement with iSchools shows the opposite. Males have a tendency to prefer sliders over Venn diagram to control and filter.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 70

Page 71: Interactive Recommender Systems

More Details on SetFusion?

• Effect of other variables: gender, age, experience with in the domain, familiarity with the system

• Check our paper in the IJHCS “User-controllable Personalization: A Case Study with SetFusion”: Controlled Laboratory study with SetFusion versus traditional ranked list

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 71

Page 72: Interactive Recommender Systems

Study 2 – UMAP 2013

• Field Study: let users freely explore the interface

- ~50% (50 users) tried the SetFusion recommender

- 28% (14 users) bookmarked at least one paper

- Users explored in average 14.9 talks and bookmarked 7.36 talks in average.

A AB ABC AC B BC C15 7 9 26 18 4 1716% 7% 9% 27% 19% 4% 18%

Distribution of bookmarks per method or combination of methods

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 72

Page 73: Interactive Recommender Systems

Hybrid RecSys: Visualizing Intersections

Clustermap Venn diagram

• Clustermap vs. Venn Diagram

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 73

Page 74: Interactive Recommender Systems

TalkExplorer vs. SetFusion

• Comparing distributions of explorations

In studies 1 and 2 over TalkExplorer we observed an important change in the distribution of explorations.

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 74

Page 75: Interactive Recommender Systems

TalkExplorer vs. SetFusion

• Comparing distributions of explorations

Comparing the field studies:- In TalkExplorer, 84% of

the explorations over intersections were performed over clusters of 1 item

- In SetFusion, was only 52%, compared to 48% (18% + 30%) of multiple intersections, diff. not statistically significant

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 75

Page 76: Interactive Recommender Systems

Summary & Conclusions

• We showed that intersections of several contexts of relevance help to discover relevant items

• The visual paradigm used can have a strong effect on user behavior: we need to keep working on visual representations that promote exploration without increasing the cognitive load over the users

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 76

Page 77: Interactive Recommender Systems

Limitations & Future Work

• Apply our approach to other domains (fusion of data sources or recommendation algorithms)

• For SetFusion, find alternatives to scale the approach to more than 3 sets, potential alternatives:– Clustering and – Radial sets

• Consider other factors that interact with the user satisfaction: – Controllability by itself vs. minimum level of accuracy

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 77

Page 78: Interactive Recommender Systems

Current Work on Interfaces

• MoodPlay– With Ivana Andjelkovic & John O’Donovan (UCSB)

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 78

Andjelkovic, I., Parra, D., & O'Donovan, J. (2016, July). Moodplay: Interactive Mood-based Music Discovery and Recommendation. In Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization (pp. 275-279). ACM.

Page 79: Interactive Recommender Systems

MoodPlay

• https://www.youtube.com/watch?v=eEdo32oOmcE

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 79

Page 80: Interactive Recommender Systems

Emotion Models

• Modelo de emociones de Russel(1980)

• GEMS (2008)

Page 81: Interactive Recommender Systems

Moods and Music: the GEMS model

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 81

Page 82: Interactive Recommender Systems

System Architecture

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 82

Page 83: Interactive Recommender Systems

Hybrid Recommendation Approach

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 83

Page 84: Interactive Recommender Systems

User Study

• Conducted on Mechanical Turk, 4 conditions

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 84

Page 85: Interactive Recommender Systems

Interactions

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 85

Page 86: Interactive Recommender Systems

Interaction Stats

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 86

Page 87: Interactive Recommender Systems

Diversity Consumption

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 87

Page 88: Interactive Recommender Systems

User Prior Mood and Artist Mood

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 88

Page 89: Interactive Recommender Systems

Post-Study Survey

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 89

Accuracy

Page 90: Interactive Recommender Systems

Post-Study Survey

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 90

Diversity

Page 91: Interactive Recommender Systems

Post-Study Survey

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 91

Confusing Interface

Page 92: Interactive Recommender Systems

Post-Study Survey

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 92

Easy to use

Page 93: Interactive Recommender Systems

CONCLUSIONS (& CURRENT) & FUTURE WORK

Page 94: Interactive Recommender Systems

Challenges in Interactive RecSys

• Objectives• Controllability• Context-aware recommendations• Privacy• Visualization Techniques• Interaction Techniques• Conversation Interfaces• Evaluation Methodology

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 94

Page 95: Interactive Recommender Systems

Future Work

• Opportunities for using new devices (Sensors on Stmartphones, EEG)

• Although new devices can capture a lot of new types of data, there is still a lot to be done with data we already produce but we haven’t consumed (user logs on social web sites, etc.)

3/29/17 D. Parra, FuturePD talk, UMAP 2016 95

Page 96: Interactive Recommender Systems

MoodPlay in the Chilean news

Page 97: Interactive Recommender Systems

MoodPlay in the Chilean news

Moodplay as therapy?

Page 98: Interactive Recommender Systems

Moodplay as therapy?

• S. Koelsch. A neuroscientific perspective on music therapy. Annals of the New York Academy of Sciences, 1169(1):374–384, 2009.

• Music can help on modulate certain mental states.

3/29/17 D. Parra, FuturePD talk, UMAP 2016 98

Page 99: Interactive Recommender Systems

Previous work: MIT Mood Meter

• http://moodmeter.media.mit.edu/

3/29/17 D. Parra, FuturePD talk, UMAP 2016 99

Page 100: Interactive Recommender Systems

Input Data: from Social Networks?

• Michelle Zhou’s personality profile

3/29/17 D. Parra, FuturePD talk, UMAP 2016 100

Page 101: Interactive Recommender Systems

Visual emotion detection

• https://github.com/auduno/clmtrackr

3/29/17 D. Parra, FuturePD talk, UMAP 2016 101

Page 102: Interactive Recommender Systems

Using EEG (BCI)

3/29/17 D. Parra, FuturePD talk, UMAP 2016 102

EMOTIVhttp://emotiv.com/epoc/

NEUROSKYhttp://neurosky.com/biosensors/

Page 103: Interactive Recommender Systems

Heatmaps to Moodplay

3/29/17 D. Parra, FuturePD talk, UMAP 2016 103

Page 105: Interactive Recommender Systems

EpistAid

• Epistemonikos: Evidence-based Medicine• Physicians answer clinical questions

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 105

Page 106: Interactive Recommender Systems

EpistAid 2

• Process of building evidence matrices is really slow

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 106

Page 107: Interactive Recommender Systems

EpistAid: IUI to support physicians

• Study 1: Relevance Feedback to find missing papers faster, off-line evaluation

• Study 2: Study with physicians at PUC

March 29th, 2017 D.Parra ~ UFMG – Invited Talk 107