Research of text paragraphs clustering strategies based on the cumulative Logistic regression analysis

Aiming at the difference between paragraphs clustering and traditional full texts clustering in useable information and clustering size, the paper proposes a new clustering strategy. It uses the idea of multiple features fusion to dig useful features as far as possible and uses the cumulative Logistic regression analysis to fit the internal relation between these features and paragraphs similarity. At last, it uses the complete-link method of hierarchical clustering to process the set of paragraphs. The results of the paragraphs similarity computation experiment and the paragraphs clustering experiment show the feasibility of the method.