Research On Key Technologies Of Ontology Learning Based On Chinese Text

With the coming the era of big data, knowledge on Internet presents exponential growth, and its content is more richer, more diverse forms. However, how to carry on the fast, accurate and effective organization and management the knowledge is becoming a new topic in the field of compute technology and its related. Ontology as a solution to share and exchange information on the basis of the semantic level, since the proposed attracted wide attention of scientists at home and abroad. Ontology and it related technologies such as ontology building,ontology mapping and ontology application research has important theoretical and practical value.In view of the current manual construction of ontology, there are many bottlenecks, such as time-consuming, difficult to dynamic update and so on. This paper for automatic or semi-automatic ontology construction, ontology learning that is carried out related research, whose main tasks include ontology concept extraction, relation extraction. The concept relation is divided into the relationship between taxonomic (hyponymy) and non-taxonomic relations. This paper carried out as the following:(1) Ontology concept is the cornerstone of the body, determine the quality of ontology building.First of all, the analysis of ontology concept extraction method of current DR&DC and CCM&TFIDF; and then, a method of concepts automatic extraction method of CCM and TFIDFE combination is proposed, finally, to make a comparative analysis of these methods to verify the proposed method effectiveness.(2)Taxonomic (hyponymy) is the basic frame body,and it can be divided into hierachical concept.The paper has combined with an improved K-means clustering algorithm,proposed a classification relation extraction method:First, constructing the domain concept vector space model VSM; then, by Euclidean distance and cosine distance combination of computing the similarity between concepts; finally, using an improved k-means algorithm to cluster concept, and the introduction of SIL index function is identified as the optimal cluster number K method, so as to achieve a better clustering effect.(3)Non-taxonomic relations are the main branches of the body, and it make the body more complete. Non-taxonomic relation extraction research for ontology learning focus and difficulty. First, The concept of co-occurrence is extracted by the method of association rules; According to the principle of the concept and the verb resonance, the VF*ICF method is used to obtain the verb and the concept of co-occurrence, and put it as a relationship label; finally, based on the log likelihood ratio method to calculate the extraction of concept and relation labels on correlation and obtain Non-taxonomic relations.Based on the above research methods, this paper has presents an ontology learning framework, and implement a prototype system of ontology learning based on text. The prototype system realizes the extraction of ontology concept, Taxonomic relationship and Non-taxonomic relationship.In the last,the extracted domain ontology is persistent to the database.