Topic estimation of web search transaction log queries using monte-carlo simulation

Abstract

A user's single session with a Web search engine may consist of seeking information on single or multiple topics. Limited research has focused on multitasking search query sessions. The objective of the study is to provide a detailed analysis of multitasking sessions and attempt to identify the topic of subsequent queries. The analysis is not only on which topics the users are interested in, but also from which topics to which topics the users are switching, hence we form topic transition matrices. Using this knowledge, Monte-Carlo simulation is used to identify the topic of upcoming queries. Findings include: (1) the number of topic shifts are small compared to the number of topic continuations in the dataset (2) the most frequently detected topics in the dataset are general information, entertainment and computers, followed by sexual, hobbies, shopping and travel in both portions of the dataset, and (3) Monte Carlo simulation and the use of conditional probabilities for subsequent queries have not performed favorably for topical estimation of subsequent queries. © 2006. Seda Ozmutlu, H. Cenk Ozmutlu and Amanda Spink.