On term selection for query expansion
Journal of Documentation
A hidden Markov model information retrieval system
Proceedings of the 22nd annual international ACM SIGIR conference on Research and development in information retrieval
Relevance based language models
Proceedings of the 24th annual international ACM SIGIR conference on Research and development in information retrieval
Model-based feedback in the language modeling approach to information retrieval
Proceedings of the tenth international conference on Information and knowledge management
Introduction to Modern Information Retrieval
Introduction to Modern Information Retrieval
Smoothing Functions for Automatic Relevance Feedback in Information Retrieval
DEXA '00 Proceedings of the 11th International Workshop on Database and Expert Systems Applications
Term feedback for information retrieval with language models
SIGIR '07 Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval
Improve retrieval accuracy for difficult queries using negative feedback
Proceedings of the sixteenth ACM conference on Conference on information and knowledge management
Selecting good expansion terms for pseudo-relevance feedback
Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval
Hi-index | 0.00 |
Traditional relevance feedback methods, which usually use the most frequent terms in the relevant documents as expansion terms to enrich the user's initial query, could help improve retrieval performance. However, in reality, many expansion terms identified in traditional approaches are indeed unrelated to the query and even harmful to the retrieval. This paper introduces a new method based on the relative word-frequency to select good expansion terms. The relative word-frequency defined in this paper is a new feature and can help discriminate relevant documents from irrelevant ones. The new approach selects good expansion terms according to the relative word-frequency and uses them to reformulate the initial query. We compare a set of existing relevance feedback methods with our proposed approach, including the representative vector space models and language models. Our experiments on several TREC collections demonstrate that retrieval effectiveness can be much improved when the proposed approach is used. Experimental results show that the improvement of our proposed approach is more than 30% over traditional relevance feedback techniques.