Research on Multi-classification and Multi-label in Text Categorization

Aiming at multi-classification and multi-label in text categorization, an apery algorithm is proposed which judges whether document has multi-classification and multi-label by estimating the similarity difference among final classifier values. If the quotient of the biggest category´s classifier value divided by the second biggest category´s classifier value is less than or equal to a threshold, the document belongs to two categories. The optimum threshold is set to 1.4 by experiment, and experiment results demonstrate performance increases by 1.42 percent.