Abstract: Thraditional clustering methods arc usually based on the similarity of keywords appearing in documents. Since these methods may lead to the loss of lots of semantic information, their clustering results are not accurate enough and often need large amount of computation. A new method for hierarchically clustering documents based on several domain ontologics was proposed. This method first transformed keyword-based vectors into corresponding concept-based vectors making use of several domain ontologies. Then, a formula was given for calculating similarities between different documents. An algorithm for document clustering based on several domain ontologics was proposed and its corresponding prime system was also designed and implemented. The experimental results show that our method can express and process documents from the perspective of concept semantics. It can decrease the amount of computation by reducing the dimension of the space of clustered o均ects and improve both the accuracy and the efficiency of document clustering.

Key words: Domain ontology, Similarity computing, Agglomcratc hierarchical clustering