Abstract：This paper presents a new algorithm using title category semantic recognition for text categorization. The algorithm generates feature space based on its category, picks up category semantic words of the title to produce candidate category and finally classifies it under these candidate categories. The experimental results firmly prove that the new algorithm performs better with fewer candidates and higher precision. Further research introduces category space representation efficiency to verify the validity of the new algorithm and proves that it can achieve great improvement in text representation.
2007, Vol. 29 
