ASSIST: Adaptive Social Support for Information Space Traversal Jill Freyne and Rosta Farzan
Problem Finding Relevant Information in existence of too much information Social Technologies Social Search Social Navigation Social Bookmarking Exploiting pools of wisdom from multiple social technologies ASSIST Exploiting Social Search and Browsing
Outline Social Search Social Navigation ASSIST Architecture ASSIST-YouTube Evaluation Plan
Social Search Re-rank and annotate search result list Using community interaction patterns Reflecting the preferences of the community Emphasizes the relationship between a page and its query terms rather than the relationship between a page and its content terms Query repetition among members of the community Example I-SPY
Social Navigation History-enriched information space Making the aggregated or individual action of others visible Reading, Annotating, watching, … Example KnowledgeSea II Footprints Augmenting the links based on number of times users are passing through a link or visiting a page considering time spent reading Annotation Augmenting the links to pages with users’ annotation
ASSIST Deployment environment ACM DL YouTube ASSIST Engine Updating search hit-matrix Updating browsing records Return recommendation using exploiting community information
Updating Search Hit-Matrix Community identification Query Related result list Page identification Considering time spent
Updating Browsing Records Simple browsing Vertical browsing to or from menu Community identification Page identification Shown items vs. browsed items Considering time spent Contextual browsing Horizontal browsing Related query Community identification, and item identification Query Distance from query Association degrades as the user browses further from the query Dividing by 2 every time (?) Related papers (video, papers) Community identification, item identification Related item identification
Recommendation Page Popular pages Pages accumulating selections from the community Indirect recommendation by adding icons Pages leading to selected page Related pages Predefined by system, e.g. related videos in YouTube Query Queries lead to selected page
Recommendation Context Search results Re-ranking search result Promoting popular result for the community Only top 3 result to allow serendipity Icon augmentation Search popularity Browsing popularity Related queries Related pages Browsing Page Icon augmentation Information Page Icon augmentation
ASSIST-YouTube – Search Result Re-ranking & Augmenting
ASSIST-YouTube – Browsing
ASSIST-YouTube – Watch Video
Evaluation Plan Targeted population 20 Graduate students at UCD Setting Voluntarily participation Setting proxy Tracking usage of YouTube for two months Social YouTube will be on and off randomly Objectives Exploring effect of social technologies in different context Exploring the values of integrating social search and browsing
Hypothesis Users’ interactions with YouTube system will be affected by added social recommendations Higher number of clicks on promoted results Higher number of clicks on augmented results Watching longer the recommended videos More successful search with social recommendations Shorter search path Ranks of the clicks
Hypothesis Integration of social search and browsing add values to the system Event based Selection probability the probability that a link accompanied by social icons will be selected Impression based selection probability the probability that a link for which the user mouses over one of its social icons will be selected. Navigation from search to browsing and vice versa
Question/Discussion