An Improved Feature Selection Method for Chinese Short Texts Clustering Based on HowNet

摘要: Short texts have played an important role in the field of text data mining. Because problems arousing from complexity Chinese semantics and sparseness, which is obvious characteristic short texts, it necessary to explore some new semantic-based methods cluster texts. An improved approach feature selection based on HowNet applied this paper address sparseness By redefining Vector Space Model semantic level merging generalized synonymy features, we present a generation strategy. Experimental results show that by similar feature, our method effective dimension reduction gets better clustering performance. The proposed HowNet-based suitable for clustering.