Wiki Map A Visual Graph of Wikipedia Articles

Wiki. Map A Visual Graph of Wikipedia Articles CSE 403, Winter 2011 Alana Killeen, Kimberly Koenig, Steven Kwan

Vision • Break out of the Wikipedia “street view”! • Allow users to browse the entire "map" of Wikipedia • An interactive, graph model for visually searching and exploring Wikipedia articles and their relationships • Connections (edges) between related articles • Zoom out/in to only see connections with high/low relevancy • Hover article “nodes” to preview content • Target = Knowledge Enthusiasts • People who love to learn and see how information is connected • Students, professionals, researchers, or casual learners • Scope • Not reinventing Wikipedia or its search engine! • Not representing all article relationships/interconnections

Software Architecture Data Tier Client Tier Logic Tier Query/Input Parser Wikipedia API Parser Page Rendering Relevancy Algorithm Front-End Wikipedia Back-End Client Tier Logic Tier Data Tier ASP. NET, HTML/XHTML, Java. Script, PHP C#, Java, PHP Wikipedia API Parse user input and queries Query Wikipedia API and parse output Communicate with Logic Tier Communicate with Client and Data Tiers Output in many formats, including: JSON, WDDX, XML, YAML, PHP’s native serialization format Render and update the UI Determine interconnected article relevancies

Challenges & Risks • Managing "information overload“ for simplicity, performance, and ideal user experience • More relationships between articles than can be represented • Which relationships to show? How many is too many? • Preserving performance with a large dataset • Mitigation strategies • Provide adjustable levels of detail/granularity (“zoom in/out”) • Different ways to determine relational strength between articles • "See also" section = high relevance • Articles mutually referencing each other (graph cycles) = high relevance • Do not attempt to preload all data at once • Render graph “on the fly” per user’s input • Predict and pre-load some data which may be useful • Higher priority to more strongly-linked relationships by default