Metrics for Assessing Sets of Subtopics

SIGIR Conf. Research and Development in Information Retrieval |

Published by Association for Computing Machinery, Inc.

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.