Research Of Text Clustering On Food Complaint Documents Based On Ontology

With the development of the Internet, the transmission of information had a huge change, people can be easily downloaded massive data from the Internet , text clustering, as an effective means of organizing text, can help people find Internet hot news, automatic edit multi-document abstract. Text clustering influenced more and more scholars' attention.The core technology of text clustering is the text representation, traditional solution is based on vector space model of presentation. This kind of statistical means are not very good expression of semantics and pragmatics text information, and therefore there are certain drawbacks, such as cannot solve the polysemy, synonyms, this also influence such as text clustering restricts the quality. In order to solve these problems, we introduce ontology technology.Ontology used to describe the nature of things, is the field of artificial intelligence, and is an emerging knowledge representation techniques. Ontology use a certain modeling language to express with people accepted empirical knowledge. it can describe concept and describe the relationship between concepts. Ontology can be considered a communication means let machine understanding of human knowledge and its positive use up to solve practical problems.This paper builds a dairy ontology aiming at the complaint documents of dairy products, and presents a frame based on ontology in text clustering, the key is to solve the expansion of semantic information on text representation, through the concept matching find concept of implicit information in the text documents, so as to added features to feature vector, finally, to improve the quality of cluster effect. Bringing topic knowledge in unsupervised clustering, in calculating similarity between document and harm topic, to prove clustering. I used this algorithm in clustering complaint text to find out useful information, tracking existing hidden trouble in the food safety. Experimental results prove that this method can finish clustering, and achieved good effect. This research has application value and broad application prospect.