Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval

Introduction. Studies have indicated that users' text highlighting behaviour can be further manipulated to improve the relevance of retrieved results. This article reports on a study that examined users’ text highlight frequency, length and users' copy-paste actions. Method. A binary votin...

Full description

Bibliographic Details
Main Authors: Balakrishnan, Vimala, Mehmood, Y., Nagappan, Y.
Format: Article
Published: Information Research 2016
Subjects:
_version_ 1796960494345519104
author Balakrishnan, Vimala
Mehmood, Y.
Nagappan, Y.
author_facet Balakrishnan, Vimala
Mehmood, Y.
Nagappan, Y.
author_sort Balakrishnan, Vimala
collection UM
description Introduction. Studies have indicated that users' text highlighting behaviour can be further manipulated to improve the relevance of retrieved results. This article reports on a study that examined users’ text highlight frequency, length and users' copy-paste actions. Method. A binary voting mechanism was employed to determine the weights for the feedback, which were then used to re-rank the original search results. A search engine prototype was built using the Communications of the ACM test collection, with the well-known BM25 acting as the baseline model. Analysis. The proposed enhanced model’s performance was evaluated using the mean average precisions and F-score metrics, and results were compared at the top 5, 10 and 15. Additionally, comparisons were also made based on the number of terms used in a query, that is single, double and triple terms. Results. The findings show that the enhanced model significantly outperformed BM25, and the rest of the models at all document levels. To be specific, the enhanced model showed significant improvements over the frequency model. Additionally, retrieval relevance was found to be the best when the query length is two. Conclusions. Users’ post-click behaviour may serve as a significant indicator of their interests, and thus can be used to improve the relevance of the retrieved results. Future studies could look into further extending this model by including other post-click behaviour such as printing or saving.
first_indexed 2024-03-06T05:45:28Z
format Article
id um.eprints-18450
institution Universiti Malaya
last_indexed 2024-03-06T05:45:28Z
publishDate 2016
publisher Information Research
record_format dspace
spelling um.eprints-184502020-01-08T07:42:10Z http://eprints.um.edu.my/18450/ Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval Balakrishnan, Vimala Mehmood, Y. Nagappan, Y. QA75 Electronic computers. Computer science Introduction. Studies have indicated that users' text highlighting behaviour can be further manipulated to improve the relevance of retrieved results. This article reports on a study that examined users’ text highlight frequency, length and users' copy-paste actions. Method. A binary voting mechanism was employed to determine the weights for the feedback, which were then used to re-rank the original search results. A search engine prototype was built using the Communications of the ACM test collection, with the well-known BM25 acting as the baseline model. Analysis. The proposed enhanced model’s performance was evaluated using the mean average precisions and F-score metrics, and results were compared at the top 5, 10 and 15. Additionally, comparisons were also made based on the number of terms used in a query, that is single, double and triple terms. Results. The findings show that the enhanced model significantly outperformed BM25, and the rest of the models at all document levels. To be specific, the enhanced model showed significant improvements over the frequency model. Additionally, retrieval relevance was found to be the best when the query length is two. Conclusions. Users’ post-click behaviour may serve as a significant indicator of their interests, and thus can be used to improve the relevance of the retrieved results. Future studies could look into further extending this model by including other post-click behaviour such as printing or saving. Information Research 2016 Article PeerReviewed Balakrishnan, Vimala and Mehmood, Y. and Nagappan, Y. (2016) Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval. Information Research, 21 (4). p. 724. ISSN 1368-1613, http://www.informationr.net/ir/21-4/paper724.html
spellingShingle QA75 Electronic computers. Computer science
Balakrishnan, Vimala
Mehmood, Y.
Nagappan, Y.
Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title_full Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title_fullStr Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title_full_unstemmed Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title_short Moving beyond text highlights: Inferring users’ interests to improve the relevance of retrieval
title_sort moving beyond text highlights inferring users interests to improve the relevance of retrieval
topic QA75 Electronic computers. Computer science
work_keys_str_mv AT balakrishnanvimala movingbeyondtexthighlightsinferringusersintereststoimprovetherelevanceofretrieval
AT mehmoody movingbeyondtexthighlightsinferringusersintereststoimprovetherelevanceofretrieval
AT nagappany movingbeyondtexthighlightsinferringusersintereststoimprovetherelevanceofretrieval