<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>5916</REFNUM><AUTHORS><AUTHOR>Chalmers,M.</AUTHOR></AUTHORS><YEAR>2001</YEAR><TITLE>Paths and Contextually Specific Recommendations</TITLE><PLACE_PUBLISHED>Proc. DELOS/NSF Workshop on Personalisation and Recommender Systems in Digital Libraries, Dublin, June 2001 </PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><LABEL>Chalmers:2001:5916</LABEL><ABSTRACT>Collaborative filtering (Goldberg 1992) or recommender systems (Resnick 1997) depend on a model of each person using them. This is the representation of the current user?s activity or information need, and the basis for personalising its provision of information. Most recommender systems collect information about each user over some time, to build up a profile of individual choices or interests. This representation is generally unordered temporally, although some approaches do focus on temporal patterns of user activity. One such approach relies on time?stamped logs of user activity??paths? through information?and represents activity or need as the recent part of the user?s path e.g. the set of URLs loaded into the user?s web browser within the last few minutes (Chalmers 1998). As in other recommenders, this path forms a model of the user that avoids content analysis, and this lets us handle complex and heterogeneous data at the cost of collecting significant data on user activity. In a way unlike most recommenders but rather like information retrieval systems, path systems allow the user to form a contextually specific representation of their current information need (Chalmers 1999). This can be done either actively or passively, but the result is a query or context that the system responds to with recommendations. In this way, path systems aim to do more than make personal or individual recommendations. They also aim to be contextually specific in their recommendations. An issue here is what one records in the path. This may be addressed by focusing on fine-grained actions such as activity within and between web pages (Villa 2001). Another approach is to focus on the media one tracks and records. As we describe below, we have enriched paths by including file activity in the xemacs editor so that, for example, recommendations are made on the basis of a combination of web pages and images browsed along with the code files recently read or edited. Recommendations can therefore be ?documents? of v</ABSTRACT><URL>http://www.dcs.gla.ac.uk/~matthew/papers/delos.html</URL></RECORD></RECORDS></XML>