Ontology Construction And Auto-Population Towards Chinese Text

Semantic Web is an important trend for the internet. One of its aims is to identify all the resources on the Web and build all sorts of machine-operable semantic relations among those resources. The Semantic Web represents information resources in an explicit and formalized manner so that it can enhance the interoperation of isomeric system and promote the sharing of information and the development of information processing technology. Ontology, as an important service layer of the Semantic Web, plays a core role in the accessing, interoperation and communication based on content.The concept of ontology is derived from philosophy research. In the field of computer science, ontology is used to describe or express a group of concepts or terms in certain domain. It can be used to organize higher level of knowledge abstraction of knowledge library and also to describe knowledge from a certain domain.Ontology building and automatic population cause more more and attension. Many researchers proposed their ontology building methods fitting for their application domain, while others began their research on how to learning ontology from text and how to extend ontology instance. But those researches are still in an immature status. This is not only related to the level of processing technology, but also due to that research of the Semantic Web has just begun and many theorical problems of ontology building still need deeper research.In this paper, based on the previous works, an ontology building method for smaller domain has been designed. With the investigation of the Chinese time description, a time ontology structure was designed and a time ontology for Chinese was built. Based on that ontology, auto-populating experiments has been carried out by making use of Support Vector Machine (SVM) to classify time instances. Both lexiccal and syntax knowledge are used to represent the features of Chinese time description. Based on the analysis of several classification strategies, a research on the construction of multi-class classifier based on SVM binary classifier was done and a multi-class classification strategy based on the system of ontology concept has been designed. Various multi-class classification strategies and kernel functions has been compared to investigate their impacts on classification effect. Experimental results indicate that our multi-class classification strategy achieved a better performance in the task of time ontology auto-population compared with other multi-class classification strategies.