关键词: 文本聚类, 文本向量, 特征选择, 位置加权, 簇间相似度

Abstract:

Document clustering is an important research topic of natural language processing and is widely applicable in the areas such as information retrieval, web mining and digital libraries. Because the feature terms of different positions in the document are different for the article’s contribution, TCABPW (a text clustering algorithm based on position weighting) is proposed in this paper. We construct a new text vector by selecting Ltopweight text that reflects the topical subject of the document and it is used to realize text clustering by hierarchical clustering and the Kmeans method. The results show that without affecting the quality of text clustering, the algorithm can not only greatly reduce the high dimension of text clustering, but also can significantly increase the stability and purity of text clutering, and can also produce the clusering effect of good quality.

Key words: text clustering;text vector;feature selecting;position weighting;similarity between clusters