Boosting Titles does not Generally Improve Retrieval Effectiveness
Author(s) -
Jimmy Jimmy,
Guido Zuccon,
Bevan Koopman
Publication year - 2016
Publication title -
qut eprints (queensland university of technology)
Language(s) - English
Resource type - Conference proceedings
DOI - 10.1145/3015022.3015028
Subject(s) - boosting (machine learning) , computer science , information retrieval , field (mathematics) , search engine , machine learning , mathematics , pure mathematics
The fields that compose structured documents such as web pages have been exploited to improve the effectiveness of in- formation retrieval systems. Field-based retrieval methods assign different levels of importance (weights) to different fields, e.g., by boosting the score of a document when query terms are found in a specific field. An important question is how to decide which field should be boosted? It has been speculated that the title field should receive a higher weight. In this paper, we investigate whether boosting the title field of structured documents actually does improve retrieval effectiveness. Our results show that, on average, boosting titles does not improve retrieval effectiveness for field-based retrieval; this is both for ad-hoc web search and exploratory- based web search tasks. However, we do find that the boosting of titles does generally improve retrieval effectiveness for navigational queries and a small subset of ad-hoc queries. This result advocates for adaptive methods that selectively adjust boosting of specific fields based on the query
Accelerating Research
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom
Address
John Eccles HouseRobert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom