Research Of Conceptual Relation Extraction Based On Topic-Text Paragraph

With the quick expanding of the Internet information resource, the number of texts increases rapidly. People hardly can find information which they need quickly. In order to solve the problem, we urgently need an automated facility to help users make use of current Internet information to gain really useful information. Research on information extraction emerges under this background. Most of the Internet information exists in the form of text which could be divided into several topic-text paragraphs. Subject concepts and concept relations extraction provides a new way for information retrieval based on paragraph topics and text automatic abstraction.The research work of this paper is a research on topic conceptual and conceptual relation extraction. In simple terms, we extract some concept which can index text-paragraph topic and construct those conceptual relations. In order to extract subject concepts, at first, by word clustering, we establish a conceptual vector space model. Based on this modal, weights of concepts can be carried out in terms of conceptual semantic similarity from hownet.Then, our paper presents a novel algorithm about choosing subject concepts based on conceptual quantified relations. We can extract more precise subject concepts. First, we can make use of Chinese Dictionary (Modern Chinese Dictionary) to gain conceptual quantified relations. Then according to conceptual quantified relations in Chinese dictionary, subject importance of concepts can be got by computing dot product of weight vectors and related vectors. At last, the subject concepts are extracted by importance.For topic concepts, we use machine learning algorithms to extract topic-conceptual relations .when words and part of speech around topic concepts were selected as feature vector, the performance of the algorithms got the peak. At last, we select some common concept relations.