Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

CSE509 Lecture 6

1,107 views

Published on

Lecture 6 of CSE509:Web Science and Technology Summer Course

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

CSE509 Lecture 6

  1. 1. CSE509: Introduction to Web Science and Technology<br />Lecture 6: Social Information Retrieval<br />ArjumandYounus<br />Web Science Research Group<br />Institute of Business Administration (IBA)<br />
  2. 2. Last Time…<br />Transition from Web 1.0 to Web 2.0<br />Social Media Characteristics<br />Part I: Theoretical Aspects<br />Social Networks as a Graph<br />Properties of Social Networks<br />Part II: Getting Hands-On Experience on Social Media Analytics<br />Twitter Data Hacks<br />Part III: Example Researches<br />August 13, 2011<br />
  3. 3. Today<br />Role of Today’s Web: Changing the way Information Needs are Satisfied<br />Social Search<br />Research Case by Microsoft Research: What do People Ask their Social Networks<br />Techniques for Influence Analysis in Social Networks<br />August 13, 2011<br />
  4. 4. Role of Today’s Web<br />August 13, 2011<br />Marketing <br />Tool<br />Information Finding Tool<br />Media <br />Tool<br />
  5. 5. New Dimensions in Search with The Social Web<br />Information Overload<br />Search engines don’t always hold answers that users are looking for<br />Smart Search (CNN Money)<br />“The Web, they say, is leaving the era of search and entering one of discovery. What’s the difference? Search is what you do when you’re looking for something. Discovery is when something wonderful that you didn’t know existed, or didn’t know how to ask for, finds you.”<br />August 13, 2011<br />What does that mean for search engines? Will they be left behind?<br />
  6. 6. Role of Today’s Web<br />August 13, 2011<br />Marketing <br />Tool<br />Information Finding Tool<br />Media <br />Tool<br />
  7. 7. Social Search<br />Takes into account the “social graph” of the person initiating the query<br />Search activity in which users pose a question to their social networks<br />Search systems using statistical analytics over traces left behind by others<br />Conducting a search over an existing database of content previously provided by other users such as searching over the collection of public Twitter posts or searching through an archive of questions and answers<br />August 13, 2011<br />
  8. 8. Social Search Benefits<br />Reduced impact of link spamby lesser reliance on link structure of Web pages<br />Increased relevance due to each result being selected by users<br />Web pages relevance judged from reader’s perspective rather than author’s perspective<br />More current results through constant feedback<br />August 13, 2011<br />Improvements achieved by social search have not been quantified so far<br />
  9. 9. What Do People Ask Social Networks?<br />Meredith Ringel Morris, MSR<br />Jaime Teevan, MSR<br />Katrina Panovich, MIT<br />August 13, 2011<br />
  10. 10. Questions about People’s Questions<br />What questions do people ask?<br />How are the questions phrased?<br />What are the question types and topics?<br />Who asks which questions and why?<br />Which questions get answered?<br />How is answer speed and utility perceived?<br />What are people’s motivations for answering?<br />August 13, 2011<br />
  11. 11. Survey of Asking via Status Messages<br />Survey content<br />Used a status message to ask a question?<br />Frequency of asking, question type, responses received<br />Provide an example<br />Answered a status message question?<br />Why or why not?<br />Provide an example<br />624 participants<br />Focus on Facebook and Twitter behavior<br />August 13, 2011<br />
  12. 12. Questions: Types<br />August 13, 2011<br />
  13. 13. Questions: Topics<br />Missing: Health, Religion<br />Politics, Dating, and Finance<br />August 13, 2011<br />
  14. 14. Questions: Who Asks What<br />August 13, 2011<br />men<br />old<br />Twitter<br />women<br />Facebook<br />young<br />
  15. 15. Questions: Motives for Asking<br />August 13, 2011<br />
  16. 16. Questions About People’s Questions<br />What questions do people ask?<br />How are the questions phrased?<br />What are the question types and topics?<br />Who asks which questions and why?<br />Which questions get answered?<br />How is answer speed and utility perceived?<br />What are people’s motivations for answering?<br />August 13, 2011<br />
  17. 17. Answers: Speed and Utility<br />94% of questions received an answer<br />Answer speed<br />A quarter in 30 minutes, almost all in a day<br />People expected faster, but satisfied with speed<br />Shorter questions got more useful responses<br />Answer utility<br />69% of responses helpful<br />August 13, 2011<br />
  18. 18. Answers: Speed and Utility<br />Fast<br />No correlation<br />Unhelpful<br />August 13, 2011<br />
  19. 19. Answers: Motives for Answering<br />Motives for Not Answering<br />- Don’t know the answer<br />- Private topic<br />- Question impersonal<br />August 13, 2011<br />
  20. 20. Answers About People’s Questions<br />The questions people ask<br />Short, directed to “anyone”<br />Subjective questions on acceptable topics<br />Social relationships important motivators<br />The questions that get answered<br />Fast, helpful responses, related to length and type<br />Answers motivated by altruism and expertise<br />August 13, 2011<br />
  21. 21. Enhancing Search using Social Network Features <br />Recency Crawling and Ranking<br />Identification of Hot Topics on Social Web [YQG+11]<br />News in the Making<br />Trend analysis<br />Event detection<br />Real-Time Search<br />Information Diffusion and Influence Analysis<br />Community Detection<br />Opinion Mining<br />August 13, 2011<br />
  22. 22. August 13, 2011<br />Nodes, Ties and Influence<br />
  23. 23. Importance of Nodes<br />Not all nodes are equally important<br />Centrality Analysis<br />Find out the most important nodes in one network<br />Commonly-used Measures<br />Degree Centrality<br />Closeness Centrality<br />Betweenness Centrality<br />Eigenvector Centrality<br />August 13, 2011<br />
  24. 24. Degree Centrality<br />The importance of a node is determined by the number of nodes adjacent to it<br />The larger the degree, the more import the node is<br />Only a small number of nodes have high degrees in many real-life networks<br />Degree Centrality<br />Normalized Degree Centrality: <br />For node 1, degree centrality is 3;<br />Normalized degree centrality is <br />3/(9-1)=3/8.<br />August 13, 2011<br />
  25. 25. Closeness Centrality<br />“Central” nodes are important, as they can reach the whole network more quickly than non-central nodes<br />Importance measured by how close a node is to other nodes<br />Average Distance<br />Closeness Centrality <br />August 13, 2011<br />
  26. 26. Closeness Centrality Example<br />Node 4 is more central than node 3<br />August 13, 2011<br />
  27. 27. Betweenness Centrality<br />Node betweenness counts the number of shortest paths that pass one node<br />Nodes with high betweenness are important in communication and information diffusion<br />Betweenness Centrality<br />The number of shortest paths between s and t<br />The number of shortest paths between s and t that pass vi<br />26<br />August 13, 2011<br />
  28. 28. Betweenness Centrality Example<br />The number of shortest paths between s and t<br />The number of shortest paths between s and t that pass vi<br />August 13, 2011<br />
  29. 29. Eigenvector Centrality<br />One’s importance is determined by his friends’<br />If one has many important friends, he should be important as well. <br />The centrality corresponds to the top eigenvector of the adjacency matrix A. <br />A variant of this eigenvector centrality is the PageRank score.<br />August 13, 2011<br />
  30. 30. Weak and Strong Ties<br />In practice, connections are not of the same strength<br />Interpersonal social networks are composed of strong ties (close friends) and weak ties (acquaintances)<br />Strong ties and weak ties play different roles for community formation and information diffusion<br />Strength of Weak Ties (Granovetter, 1973) <br />Occasional encounters with distant acquaintances can provide important information about new opportunities for job search<br />August 13, 2011<br />
  31. 31. Connections in Social Media<br /><ul><li>Social Media allows users to connect to each other more easily than ever</li></ul>One user might have thousands of friends online<br />Who are the most important ones among your 300 Facebook friends?<br /><ul><li>Imperative to estimate the strengths of ties for advanced analysis </li></ul>Analyze network topology<br />Learn from User Profiles and Attributes<br />August 13, 2011<br />
  32. 32. Learning from Network Topology<br />Bridges connecting two different communities are weak ties<br />An edge is a bridge if its removal results in disconnection of its terminal nodes<br />e(2,5) is a bridge<br />e(2,5) is NOT a bridge<br />August 13, 2011<br />
  33. 33. “shortcut” Bridge<br />Bridges are rare in real-life networks<br />Alternatively, one can relax the definition by checking if the distance between two terminal nodes increases if the edge is removed<br />The larger the distance, the weaker the tie is<br />d(2,5) = 4 if e(2,5) is removed<br />d(5,6) = 2 if e(5,6) is removed<br />e(5,6) is a stronger tie than e(2,5)<br />August 13, 2011<br />
  34. 34. Neighborhood Overlap<br />Tie Strength can be measured based on neighborhood overlap; the larger the overlap, the stronger the tie is<br />-2 in the denominator is to exclude vi and vj<br />August 13, 2011<br />
  35. 35. Neighborhood Overlap<br />Tie Strength can be measured based on neighborhood overlap; the larger the overlap, the stronger the tie is<br />-2 in the denominator is to exclude vi and vj<br />August 13, 2011<br />
  36. 36. Learning from Profiles and Interactions<br />Twitter: one can follow others without followee’s confirmation<br />The real friendship network is determined by the frequency two users talk to each other, rather than the follower-followee network<br />The real friendship network is more influential in driving Twitter usage<br />Strengths of ties can be predicted accurately based on various information from Facebook<br />Friend-initiated posts, message exchanged in wall post, number of mutual friends, etc. <br />Learning numeric link strength by maximum likelihood estimation<br />User profile similarity determines the strength<br />Link strength in turn determines user interaction<br />Maximize the likelihood based on observed profiles and interactions<br />35<br />August 13, 2011<br />

×