Output details
11 - Computer Science and Informatics
University of Edinburgh
Fast query expansion using approximations of relevance models
<17> Originality: We present an algorithm that speeds up the runtime of relevance models by more than two orders of magnitude. The algorithm is based on auxiliary structures, created at indexing time. We explore randomised techniques and distributed architectures for creating these structures efficiently.
Significance: Relevance models give excellent retrieval accuracy, but are very slow. This limits their application in many domains. Our algorithm reduces the run-time by 99% and allows applications on web-scale corpora.
Rigour: Published in ACM CIKM -- an international conference with a 17% acceptance rate and a high impact factor (#3 worldwide in Information Retrieval).