Skip to Main Content
Accurately estimating language model is important to improve the performance of information retrieval. The key problems include solving synonymy and polysemy problem, and smoothing the seen term or not seen term in a document. In this paper, we propose a new method for topic language model. First, concept-based clustering is performed using improved fuzzy c-means. The clustering result is considered as the topics of document collections. The probability of a document generating the topics is estimated by the similarity between the document and each cluster. Then, the probability of the topics generating words is estimated using Expectation Maximization algorithm. At last, we integrate the above algorithms into aspect model to form our topic language model. This new language model accurately describes the distribution probability of the words in different topics and the probability of a document generating a topic. Moreover, it can solve synonymy and polysemy problems. The new method is evaluated on TREC 2004/05 Genomics Track collections. Experiments show that the retrieval performance is greatly improved by the new method compared with the simple language model.
Date of Conference: 21-23 May 2007