Combining evidence for automatic Web session identification

• Published in: Information processing & management Vol. 38; no. 5; pp. 727 - 742
• Main Authors: He, Daqing, Göker, Ayşe, Harper, David J
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 2002; Elsevier Science Ltd
• Subjects: Context; Contextual Analysis; Dempster–Shafer theory; Evaluation; Information Needs; Information retrieval; Internet; Retrieval performance measures; Search context; Search Strategies; Searches; Searching; Session identification; Studies; Subject Index Terms; Transaction Log Analysis; User Needs (Information); Users; Web user logs; World Wide Web; Web user logs; Dempster–Shafer theory; Search context; Session identification
• ISSN: 0306-4573; 1873-5371
• DOI: 10.1016/S0306-4573(01)00060-7

Contextual information provides an important basis for identifying and understanding users' information needs. Our previous work in traditional information retrieval systems has shown how using contextual information could improve retrieval performance. With the vast quantity and variety of information available on the Web, and the short query lengths within Web searches, it becomes even more crucial that appropriate contextual information is extracted to facilitate personalized services. However, finding users' contextual information is not straightforward, especially in the Web search environment where less is known about the individual users. In this paper, we will present an approach that has significant potential for studying Web users' search contexts. The approach automatically groups a user's consecutive search activities on the same search topic into one session. It uses Dempster–Shafer theory to combine evidence extracted from two sources, each of which is based on the statistical data from Web search logs. The evaluation we have performed demonstrates that our approach has achieved a significant improvement over previous methods of session identification.