Text Clustering Algorithm Based on the Graph Structures of Semantic Word Co-occurrence

摘要

Text theme is the key of text clustering, while the co-occurrence words can be very stronger to express text theme in document. This paper proposes a text clustering algorithm based on the text semantic representation and the graph structure of word co-occurrence on the basis of in-depth studying text theme mining and word co-occurrence. First, the algorithm constructs the text graph-structure according to the co-occurrence of feature words. In other words, it uses the graph structure to represent all texts. Then, it adopts the maximum common sub-graph between two texts to calculate their similarity and combines with K-means clustering algorithm to realize the document clustering. The compared experimental results with hierarchical clustering algorithm show the K-means clustering algorithm based on the graph structures of word co-occurrence greatly reduce the high dimension of text vector and the algorithm complexity, significantly improves the efficiency and accuracy of text clustering, and it can also produce the clustering effect of good quality.