Constructing term thesaurus using text association rule mining

This paper presents a new algorithm called ldquoconcept-groupingrdquo that adapts an association rule mining technique to construct term thesaurus for data preprocessing purpose. Similar terms, which are written differently, can be grouped together into the same concept based on their associations before they are used for subsequent analysis. This data preprocessing is important since it has an impact on the quality of other data mining techniques such as data clustering. The algorithm is applied to bibliographic databases such as INSPEC and EI Compendex toward the objective of enhancing traditional bibliometrics and content analysis. From the experiments with a set of publication abstracts, applying the proposed algorithm to combine similar terms into a pertinent concept before clustering process yields better cluster quality.