Metrics for Assessing Sets of Subtopics

Filip Radlinski, Martin Szummer, and Nick Craswell

Abstract

To evaluate the diversity of search results, test collections have been developed that identify multiple intents for each query. Intents are the different meanings or facets that should be covered in a search results list. This means that topic development involves proposing a set of intents for each query. We propose four measurable properties of query-to-intent mappings, allowing for more principled topic development for such test collections.

Details

Publication typeInproceedings
Published inSIGIR Conf. Research and Development in Information Retrieval
Pages853-854
PublisherAssociation for Computing Machinery, Inc.
> Publications > Metrics for Assessing Sets of Subtopics