关键词: 短文本, 文本相似度, 动态表示向量, 文本聚类, K-means算法

Abstract: Since Chinese short text is short of keywords and full of anomalous words，it brings about short text similarity drift and the traditional text clustering method is not directly suitable for short text clustering.To solve the problem of sparse key-words and similarity drift in short text segments，this paper proposes a new method to build dynamic text vector by text similarity based on HowNet.This method can measure the similarity between short text segments by dynamic text vector，then find short text relationship，so as to relieve these two characteristics’ bad influence on the clustering performance and therefore to gain a better clustering result.Experiments show the method can get better performance in Chinese short text clustering，compared with traditional method.

Key words: short text, similarity between short text segments, dynamic vector, text clustering algorithm, K-means algorithm