| Ontology is located in the core layer of the Semantic Web, which is known as the next generation Internet. Therefore, the study of ontology construction method is very important. Because ontology is the mapping of the whole real world, the direct study on the method of ontology construction is impossible. So, ontology could be divided into domain ontology based on the idea of divide and conquer, changing this problem into the research of domain ontology construction method. There are lots of different characteristics between Chinese and other languages, especially the Indo-European languages. Thus, the ontology construction method based on the English, which is studied most, is difficult to be directly used for Chinese domain ontology construction. Methods of Chinese domain ontology construction are much less than others’. First of all, the choice of corpora is a big difficulty. Hence, there are a lot of methods of the domain concepts extraction. Baidu baike has innate advantages in determination of domain concepts. It is an encyclopedia dictionary edited in the way of community type based on the wiki mode, which has the characteristics of dictionary and features of wiki. These features are particularly suitable for the construction of ontology. The choosing of Baidu baike as the corpora aims to obtain the benefits of wiki edit mode that the items can be widely accepted and gets the superiorities on ontology construction of semi-structured dictionary corpus.A Chinese domain ontology construction method based on Baidu baike is proposed in this master’s thesis. Cyclic Acquisition Process approach, the Top-Down and Bottom-Up three principal methods are combined in this method and the design of that we take concept as the core for the inner loop, the ontology as the core for outer loop is adopted. This method supports the circulation mechanism of updating and improving. In this method, we propose three major principles for building ontology based on Baidu baike, and design a method for looking for the hypernym according to the characteristics of the formal language in Chinese with priority based on the three principles; According to the features of Baidu baike, the TF_IDF similarity calculation method is improved, and we propose the improved algorithm in several applications of ontology construction. Finally, we test the "rosaceae" domain ontology construction, and the experiment results are analyzed. |
