Output details
11 - Computer Science and Informatics
University of Edinburgh
A generative theory of relevance
<17> Originality: The book presents a novel way of modelling relevance for the field of Information Retrieval (IR). The model overcomes the limitations of two dominant approaches to IR: the Probabilistic Model and the Language-Modelling Framework.
Significance: The model provides a uniform way of approaching a number of search tasks, including ad-hoc retrieval, cross-language search, image annotation and retrieval, handwriting retrieval and semi-structured search with missing data. The model is formal and effective, showing significant improvements over state-of-the-art (as of 2008) approaches on a number of datasets.
Rigour: The book received positive reviews from three recognised experts in the field.