Improvement of Retrieval Efficiency using Clustered Base Query Expansion

Abstract

There has been an increasing numbers of electronic document distributed online in nowadays. Most of the data available is in a semi-structure and unstructured format. To make use of these data, we need to understand the underlying structure at some level in order for data becomes useful. This can be done using text pre-processing process and data structure analysis with natural language processing process. In information retrieval, interested documents can be retrieved using indexing technique consisting of a list of sorted terms that link to a list of documents. Each document is associated with the frequency count indicating the number of times that a term appear in the document. The users can then search for the interested documents by specify one or more keywords called ‘terms query’. These keywords can be either a single word or multiple words together. The key issue in identifying search terms is that most users may not be able to identify the search terms correctly. The problems occur when user do not know what terms are indexed. Therefore, the interested documents can not be found. This paper proposes a technique that expand the query terms based on the term similarity and terms with similar meaning within the same domain. By measuring the value of recall and precision, this technique has proven that this technique resulted in increased document retrieval performance.