Research on Food Complains Document Classification Based-on Topic

In this paper, we design a classifier based-on topic for food complain documents, and take a series of measures to the implementation process. In order to accomplish feature reduction, the filter method named term filtering for independent topic features is proposed to compress each topic feature vector. We introduce the created food ontology as background knowledge and to expand the semantic of complaint documents with the aid of HowNet. So we can supplement effective information and improve the effect of text classification. In addition, we take account of different importance between title and body in the text, considering that title can stand out the topic of text better than the textual body. Consequently, we separately calculate the word frequency which words are in textual title and body. The experiments show that it is necessary to consider the different importance between textual title and body, and we can achieve good feature reduction effect using the proposed filter method, and the classification performance get obvious improvement after the process of term expanding.