2003-08-29 Common web paths in group adaptive system http://www.ht03.org/papers/pdfs/36.pdf Abstract from http://www.ht03.org/papers/ "In this paper we describe the how we use users' accesses and interactions on pages to discover and recommend relevant Common Paths to a group of users. We collect data using a social navigation environment GAS (Group Adaptive System) we developed [1] and we are currently integrating the 'Common Path' navigation tool in the user interface. The goal is to use the Common Path of a subset of users in the system as a recommendation to a user (not in the subset)" presentation: . cooperation on the www . GAS: a group adaptive system . common web pahts in GAS web is not a social place... - everyone navigates in isolation - wikis etc are a good attempt at social stuff, but not navigation - absolutely different from the real life - we want a social nav place where pages can be scored, sent messages about, annotated by a group of people - ...where we can get knowledge of whole group other systems: - GroupLens - Fab (collab filtering for movies and web) - Swiki (server-side annotation) - Babbles (chat annotation/awareness) GAS: - share effort of finding stuff - access common info space they used the group interest to suggest/recommend stuff to the group, using the Kleinberg algorithm (HITS) [HITS has come up a lot. i should learn about the *information retrieval* field] GAS is implemented as a web proxy. how to define common paths: * longest common subsequence (LCS) appears too restrictive - if the navigations are close but not identical, eg . abcdefg vs axbyczdj * LCSs do not capture personal nature of navigation, train of thought etc * "Levenshtein distance" -- allow a bounded number of mismatches [oooh!] common paths are interesting because similarities should common interest, but also: - same reaction to information on the page - same order to access information applications: - identify common paths and suggest next steps - characterise the group and let users compare themselves to that