Improving web retrieval precision based on semantic relationships and proximity of query keywords

Authors:
Chi Tian;Taro Tezuka;Satoshi Oyama;Keishi Tajima;Katsumi Tanaka
Affiliations:
Informatics of the Faculty of Engineering, Kyoto University, Kyoto, Japan;Department of Social Informatics, Graduate School of Informatics, Kyoto University, Kyoto, Japan;Department of Social Informatics, Graduate School of Informatics, Kyoto University, Kyoto, Japan;Department of Social Informatics, Graduate School of Informatics, Kyoto University, Kyoto, Japan;Department of Social Informatics, Graduate School of Informatics, Kyoto University, Kyoto, Japan
Venue:
DEXA'06 Proceedings of the 17th international conference on Database and Expert Systems Applications
Year:
2006

Citing 3
Cited 2

Passage-level evidence in document retrieval

SIGIR '94 Proceedings of the 17th annual international ACM SIGIR conference on Research and development in information retrieval
Real life information retrieval: a study of user queries on the Web

ACM SIGIR Forum
Information Retrieval

Information Retrieval

Proximity Scoring Using Sentence-Based Inverted Index for Practical Full-Text Search

ECDL '08 Proceedings of the 12th European conference on Research and Advanced Technology for Digital Libraries
GA on IR: Study the Effectiveness of the Developed Fitness Function on IR

International Journal of Artificial Life Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

Based on recent studies, the most common queries in Web searches involve one or two keywords. While most Web search engines perform very well for a single-keyword query, their precisions is not as good for queries involving two or more keywords. Search results often contain a large number of pages that are only weakly relevant to either of the keywords. One solution is to focus on the proximity of keywords in the search results. Filtering keywords by semantic relationships could also be used. We developed a method to improve the precison of Web retrieval based on the semantic relationships between and proximity of keywords for two-keyword queries. We have implemented a system that re-ranks Web search results based on three measures: first-appearance term distance, minimum term distance, and local appearance density. Furthermore, the system enables the user to assign weights to the new rank and original ranks so that the result can be presented in order of the combined rank. We built a prototype user interface in which the user can dynamically change the weights on two different ranks. The result of the experiment showed that our method improves the precision of Web search results for two-keyword queries.