Slicing and dicing the information space using local contexts

摘要

In recent years there has been growing interest in faceted grouping of documents for Interactive Information Retrieval (IIR). It is suggested that faceted grouping can offer a flexible way of browsing a collection compared to clustering. However, the success of faceted grouping seems to rely on sufficient knowledge of collection structure. In this paper we propose an approach based on the local contexts of query terms, which is inspired by the interaction of faceted search and browsing. The use of local contexts is appealing since it requires less knowledge of the collection than existing approaches. A task-based user study was carried out to investigate the effectiveness of our interface in varied complexity. The results suggest that the local contexts can be exploited as the source of search result browsing in IIR, and that our interface appears to facilitate different aspects of search process over the task complexity. The implication of the evaluation methodology using high complexity tasks isalso discussed.