Improving Web Information Retrieval Using Shared Contexts

Andy M. Connor; Dilip K. Limbu; Stephen G. MacDonell; Russel Pears
The effective utilisation of a user's context in improving the performance of web search engines is a subject of intense research interest. In particular, much attention has been directed to the enhancement of queries and the provision of more relevant information by taking user context into account. Progress in this field has been limited to date, however, due to ongoing challenges in capturing and representing contextual information. We describe here the development and evaluation of a web-based contextual information retrieval that addresses some of these challenges and makes progress in defining the information required to create contextual profiles. Our system collects and leverages implicit and explicit user data to modify queries with the aim of more accurately reflecting the user's interests. This data is maintained dynamically in each user's contextual profile and utilised to improve the quality of information found during web searches. Where enabled, this data also contributes to the development of a shared contextual knowledge base that can also be used to augment queries. This shared contextual knoweldge base is a key aspect of this research. The system has been tested in an observational study that has considered its ability to improve the user's web search experience. This paper presents experimental data to provide evidence of the system's performance, demonstrating that the shared contextual knowledge base extends the functionality associated with the individual contextual profile.