Research On Ontology-based Short-text Classification

With the development of the Internet, data and information have become a mass. Text classification is the key method for processing and organizing large number of text, which is convenient to find precise knowledge according to needs. Traditional text classification methods are focused on normal text (long text) classification, however, short text is widely applied in the real world, such as web search clips, forums and chat messages, news feeds of blog, summary of the book or film, products introduction, user evaluation, and so on. The traditional methods of short text classification are similarity measure or web-based kernel function method. Due to the characteristics of a short text which led to sparsity of the matrix, the effect of classification did not achieve satisfactory accuracy. With the explosive growth, sthort text contains so much rich information and is very interested. However, the sparsity characteristics of short text increase the difficulty of study and make high demanding. Therefore, short text classification is very difficult and one of the hot research.This paper presents a framework of ontology-based short text classification, which focuses on how to make the information of sparsity documents more rich. And how to mine useful information in order to make characteristics better the short text classification l. The experimental results show that the method in this paper can be implemented and achieve better results.Ontology is a method as a knowledge organization and knowledge representation, which have concept hierarchy and logical reasoning.It can be express semantic from relation between the concepts. Without training samples when using this method, we can get semantic information of ontology and combine the similarity calculations to achieve the short-text classification. This study has great practical value and a wide application prospects.