Facets provide an efficient way to analyze and navigate the search result space. However, we believe facet selection has been guided by suboptimal facet and facet term properties. In this paper, we present features that rank facets based on their utility to partition the search result documents. Moreover, we propose a computationally inexpensive facet generation algorithm, provide a novel approach to extract term lists from HTML documents, and analyze facet feedback models using BM25F as baseline retrieval model. We show that the proposed approach outperforms existing algorithms and discuss the significantly reduced computational costs.
展开▼