Presentation is loading. Please wait.

Presentation is loading. Please wait.

Computational User Modeling in Search Engine Logs Hongning Wang Department of Computer Science, University of Illinois at Urbana-Champaign Urbana, IL 61801.

Similar presentations


Presentation on theme: "Computational User Modeling in Search Engine Logs Hongning Wang Department of Computer Science, University of Illinois at Urbana-Champaign Urbana, IL 61801."— Presentation transcript:

1 Computational User Modeling in Search Engine Logs Hongning Wang Department of Computer Science, University of Illinois at Urbana-Champaign Urbana, IL 61801 USA

2 User understanding via search logs mining UserQueryDocumentsClicks sochi winter Olympics obamacare affordable health care plan super bowl 2014 sochi winter Olympics health care reform Query-centric analysis: Query categories [Jansen et al. IPM 2000] Temporal query dynamics [Kulkarni et al. WSDM’11] 1.Isolated analysis 2.Holistic view Click-centric analysis: Interpreting clickthrough data [Joachims, et al. SIGIR’05, Agichtein, et al. SIGIR’06] Click modeling [Dupret and Piwowarski SIGIR’08, chalell and Zhang WWW’09]

3 Latent user group: a homogenous unit of query and clicks f1f1 Group k f2f2 p(Q) q1q1 q2q2 q3q3 Modeling of search interest Modeling of result preferences

4 User: a heterogeneous mixture over the latent user groups … Group 1 Group 2 Group k BM25 Nutrition of fruits f2f2 p(Q) q1q1 q2q2 q3q3 Stock market BM25 f2f2 q1q1 q2q2 q3q3 p(Q) Stock market market report AAPL apple TWTR apple orange banana nutrition fruit receipt fidelity online login fruit smoothie FB GOOG BM25 PageRank

5 Generation of latent user groups: Dirichlet Process priors [Ferguson, 1973] …… f1f1 Group kf2f2 p(Q) q1q1 q2q2 q3q3 Group 1 f1f1 f2f2 q1q1 q2q2 q3q3 p(Q) f1f1 Group cf2f2 p(Q) q1q1 q2q2 q3q3

6 Another layer of DP to support infinite mixture of latent user groups [Teh et al., 2006] Group 1 f1f1 f2f2 f1f1 Group kf2f2 …… Group 1 Group 2 Group k Group 1 Group 2 Group k Group 1 Group 2 Group k p(Q) q1q1 q2q2 q3q3 q1q1 q2q2 q3q3 …… f1f1 Group cf2f2 p(Q) q1q1 q2q2 q3q3 … … …

7 Query distribution in latent user groups breaking news events entertainment sports celebrities country names Group Top Ranked Queries 1iran, china, libya, vietnam, syria 2selena gomez, lady gaga, britney spears, jennifer aniston, taylor swift 3fake tupac story, pbs hackers, alaska earthquake, southwest pilot, arizona wildfires 4joplin missing, apple icloud, sony hackers, google subpoena, ford transmission 5casey anthony trial, casey anthony jurors, casey anthony, crude oil prices, air france flight 447 6tree of life, game of thrones, sonic the hedgehog, world of warcraft, mtv awards 2011 7the titanic, the bachelorette, cars 2, hangover 2, the voice 8los angeles lakers, arsenal football, the dark knight rises, transformers 3, manchester united 9miami heat, los angeles lakers, liverpool football club, arsenal football, nfl lockout 10today in history, nascar 2011 schedule, today history, this day in history

8 Click preferences in latent user groups breaking news events entertainment sports celebrities country names document age query match in title proximity in title site authority “today in history” Global model

9 Research summary User Modeling in Search Logs via A Non-parametric Bayesian Approach [WSDM’14a] Personalized Ranking Model Adaptation for Web Search [SIGIR’13] Adapting Deep RankNet for Personalized Search [WSDM’14b] Joint Learning Approach from Clickthroughs [book chapter] Research summary: http://sifaka.cs.uiuc.edu/~wang296/competition2013.html


Download ppt "Computational User Modeling in Search Engine Logs Hongning Wang Department of Computer Science, University of Illinois at Urbana-Champaign Urbana, IL 61801."

Similar presentations


Ads by Google