Improving Web Search by Query Expansion with a Small Number of Terms
Tomonari Masada, Teruhito Kanazawa, Atsuhiro Takasu, Jun Adachi · 2005
This paper presents comprehensive experimental results on query expansion (QE) with a small number of terms. Our retrieval process uses a term weighting scheme based on a probabilistic model. The terms of large Robertson’s selection values are used for QE. In most QE experiments, hundreds of terms are added to an original query. However, this style of QE substan-tially increases search response time. In this paper, we provide experimental results attained by using at most 20 terms for expansion. The best average precision is nearly 8 % increase over the baseline case, i.e., the case where we do not use QE. This improvement is ob-tained when only ten terms are added to an original query. Our results show that we can achieve a fairly good improvement even with a small number of expan-sion terms. The results also show how we should ad-just parameter values to improve the quality of seach systems using QE.