A New Semantic-Based Hierarchical Clustering For Textual Data Algorithm

The suffix tree clustering (STC) is a fast incremental clustering and it uses a tree form as a structure to represent data. However, hierarchy of clusters that are generated by taking the hierarchy of the constructed suffix tree directly is not reasonable. That reason is meaning of the words are ignored and some large-sized with poor quality of clusters are returned by STC. In this paper, a new semantic-based hierarchical clustering for textual data is proposed. The propose uses semantic suffix net as a structure to represent data and net pruning techniques as a logic to combine the related suffixes through their suffix links. Consequently, the hierarchy of clusters is generated per se when the resulting of tree pruning techniques is a tree form. While textual information can be group by using both string matching and meaning of the words as well as the hierarchy of their clusters can be return directly. That means the quality of clusters is increasing. Also, the merge or split suffered of a hierarchical clustering is reduced. Therefore, the propose algorithm can be used to reduce the suffered of a hierarchical clustering and enhance the quality of clusters.