An approach for the capture of context-dependent document relationships extracted from Bayesian analysis of users' interactions with information

A number of technologies exist which enable the unobtrusive capture of computer interface interactions in the background of a user's working environment. The resulting data can be used in a variety of ways to model aspects of search activity and the general use of electronic documents in normal working routines. In this paper we present an approach for using captured data to identify relationships between documents used by an individual or group, representing their value in a given context—that may relate to specific information need or activity. The approach employs the use of a naïve Bayesian classifier to evaluate possible relationships that are derived implicitly from the data. It is intended that the relationships established be stored within an information retrieval (IR) system to aid in the retrieval of related documents where future users arrive at a similar context. In the evaluation of the approach over 70 hours of data from computer users in industrial and academic settings are collected to assess its overall feasibility. The results indicate that the approach provides a useful method for the establishment of identifiable relationships between documents based on the context of their usage, rather than their content.