tasteweights: visual interactive hybrid recommendations
DESCRIPTION
Slides from ACM Recommender Systems conference, September 2012. Presented by John O'DonovanTRANSCRIPT
![Page 1: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/1.jpg)
TasteWeightsA Visual Interactive Hybrid Recommender System
Svetlin Bostandjiev, John O’Donovan, Tobias HöllererDepartment of Computer Science
University of California, Santa Barbara{alex, jod, holl}@cs.ucsb.edu
ACM Recommender Systems. Burlington Hotel, Dublin. September 10th 2012
![Page 2: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/2.jpg)
MotivationProblems• Many traditional
recommenders are “black boxes” and lack explanation and control [Herlocker]
• “Why am I being recommended this movie? I don’t like horror films.”
• Even in modern recommenders, data can be static, outdated or simply irrelevant from the beginning.
• Data about users and items is spread far and wide.
![Page 3: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/3.jpg)
MotivationChallenges: Need for more dynamic, more
adaptable algorithms that can cope with diverse data from APIs.
And, we need an interface that can keep up…
Solutions?User interfaces help to explain provenance of a recommendation. This can improve users’ understanding of the underlying system and contribute to better user experience and greater satisfaction
Interaction allows users to: tweak otherwise hidden systems settings; provide updated preference data, recommendation feedback etc. etc.
![Page 4: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/4.jpg)
TasteWeights: BackgroundInitial Work on Graph-based
Representations of Collaborative Filtering Algorithms:PeerChooser : Based on static MovieLens dataSmallWorlds: Web-based, dynamic data from Facebook API.
Issues discovered during evaluations:PeerChooser: Interaction with nodes that represent movie genres …too coarse.
SmallWorlds: A “complete” representation, but far too complicated view.
Learning from evaluations:Abstraction, Detail-on-demand, Interactive Visual Cues, Cleaner game-like graphics, and more flexible API connectivity. Focus on “social” recommendation.
![Page 5: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/5.jpg)
![Page 6: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/6.jpg)
Supports user interaction to update information at recommendation time
>Solves stale data problem.
Makes the ACF algorithm transparent and understandable.
>increases satisfaction, acceptance etc.
Enables fast visual exploration of the data
>what-if scenarios>increases learning
Interactive, Trust-based Recommender for Facebook Data
![Page 7: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/7.jpg)
Combining Social and Semantic Recommendations
Facebook / Twitter (Social
Recs)
DBPedia/Freebase
(Semantic Recs)
![Page 8: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/8.jpg)
TasteWeights Design
![Page 10: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/10.jpg)
ApproachParallel hybrid recommender system
InputData(FB likes)
Twitter Sources
(WF Experts)
Facebook Sources
(band pages)
WikipediaSources(W articles)
OutputData
(recs)
Input DataResolution(W articles)
SimilarityModel
RecAlgorithm
(recs)
HybridRec
Algorithm
Input DataResolution(FB pages)
Input DataResolution(TW #tags)
SimilarityModel
RecAlgorithm
(recs)
RecAlgorithm
(recs)
SimilarityModel
Entity Resolution Recommendation Hybridization
![Page 11: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/11.jpg)
Recommendation SourcesInput Data Resolution
Mapping between Wikipedia articlesFacebook pages, and Twitter #tags
Similarity ModelsWikipedia
(Data source: DBpedia)
Facebook(Data source: Facebook Graph API)
Twitter(Data source: wefollow.com)
![Page 12: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/12.jpg)
Generating Recs.
Individual Source
Hybrid StrategiesWeighted
Mixed
Cross-source
![Page 13: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/13.jpg)
Evaluation
GoalsEvaluate combining social and semantic recommendationsEvaluate explanation and transparency in a hybrid recommenderEvaluate interaction in a hybrid recommender
SetupSupervised user study. 32 participants from the human subject pool at UCSB
ProcedurePre-questionnaireTasks
Interact with ProfileInteract with SourcesInteract with Full interfaceRate recommendations
Post-questionnaire (Explanation & Interaction)
![Page 14: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/14.jpg)
Evaluation: Accuracy
ExperimentOne-way repeated measures ANOVACompared 9 recommendation methods (below) in terms of rec. accuracy
Method (independent variable)Single-source: Wikipedia, Facebook, TwitterHybrid: Weighted, Mixed, Cross-sourceInteraction: Profile, Sources, Full
Accuracy (dependent variable)Measured in terms of “Utility”
![Page 15: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/15.jpg)
Results: Accuracy
Plot of means of recommendation methods over utility with 95% confidence intervals
Results from a Tukey post-hoc analysis of the recommendation methods: multiple comparisons of means with 95% family-wise confidence level
![Page 16: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/16.jpg)
![Page 17: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/17.jpg)
![Page 18: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/18.jpg)
Results: Diversity
1000 10000 100000 1000000 10000000 1000000000
20
40
60
80
100
120
140
Diversity
Wikipeda Facebook (CF)
Hybrid Hybrid + Interaction
Bins for Number of Facebook Likes
Num
ber o
f Rec
omm
ende
d Ite
ms
![Page 19: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/19.jpg)
TasteWeights on LinkedIn DataCase Study: Portability of TW interface
-Developed a Social-Semantic Recommendation algorithm for data from LinkedIn API -Personalized for one “active” logged-in user.
-Visualized the algorithm in TasteWeights interface
Algorithm: -Map profile items to noun-phrases -Resolve to Wikipedia articles -e.g.: ph.D => PHD, UCSB => UC Santa Barbara -Compute similarity based on overlap in resolved entities.
Features-Segmented / Organized user profile-Interactive profile weighting-Interactive weighting of social connections-Dynamic re-ranking of recommendations (visual feedback)-Provenance views to show effects of each interaction.
![Page 20: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/20.jpg)
![Page 21: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/21.jpg)
ConclusionsUI and interaction design are important considerations for RSs
-Increased explanation, provenance-Expose otherwise hidden controls (e.g: control of hybrid recommender)-Helps ease the stale data problem-Support user input at various granularity (recommended item, recommendation partner, profile items etc)
-Increase ambient learning.-Promote interest in the recommender system (game-like feel)
Contributions: -Demonstrated a novel interactive RS -Hybrid of recommendations from Wikipedia, Facebook and Twitter -Evaluated via a 32 person supervised user study at UCSB. -Demonstrated portability of the system on LinkedIn’s API.
Results-Interaction increases user satisfaction in all conditions. (more interaction = higher accuracy)-Cross-source hybrid strategy outperformed individual source strategies.
![Page 22: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/22.jpg)
After the break… Inspectability and Control in Social Recommenders
In this work we touched on the ideas of inspectability and control in the context of our hybrid recommender system.
In the next talk, Bart Knijnenburg (UC Irvine) will present results from a larger study that focuses on a general analysis of inspectability and control in social recommenders. This study used some components from our TasteWeights system.
![Page 23: TasteWeights: Visual Interactive Hybrid Recommendations](https://reader037.vdocument.in/reader037/viewer/2022102823/54997d1eb47959114a8b4582/html5/thumbnails/23.jpg)
Thanks for listening!