| As the rapid development of the Internet, it is difficult for people to find whatthey really want from a mass of data. The search engine provides an effective way forlocating the useful information in enormous Web sties. For example, the Yahoo,Google and Baidu all are popular search engine widely used. For current searchengine, the retrieval mechanism is mainly based on keywords. Therefore there aremuch redundant and undesired information in the returned result, finally with lowoverall precision. So it becomes a hot pursuit to find a effective way to acquirerelevant information. At the same time, the research on network-oriented automaticQuestion-Answering (Q&A) system is paid more attention. Because it not onlyprovides a friendly, familiar human-machine interface for user, but also present ahigher precision than traditional information retrieval technology (IR).Q&A system is an application of natural language processing (NLP), includingmuch base technology of NLP and IR, such as lexical analysis, parsing, semanticanalysis, indexing, retrieval and ranking. Q&A system is composed of threecomponents - query analysis, information retrieval and answer extraction. In thispaper, we research on Web-Oriented Chinese Q&A System Relative Technologies andimplement a Chinese Q&A system, including:(1)Shallow semantic parsing. We use the method of machine learning forsemantic role labeling, and then build semantic frame structure. (2)Multi-strategy retrieval method. We build knowledge base of definition,FAQ knowledge base, local retrieval and web retrieval for different retrieval purpose. (3)Semantic frame structure matching. Answering extraction depends onsemantic frame structure matching which is used to measure the similarity betweenquery and candidate document sentences.Semantic frame structure is the key of this paper, and it pays a key role in queryanalysis, information retrieval and answering extraction. Furthermore, it establishesthe foundation structure for deep semantic analysis. |
PDF Full Text Request