Design And Implementation Of A Question Answering System For The Military Information Domain Based On The Semantic Web Technology

As an advanced form of information retrieval,automatic question answering is a method to obtain accurate answers directly.Semantic Web technology constantly makes web data opening and structured to promote sharing and integration of data,so that machines can understand and manipulate data more easily.Semantic Web technology enables traditional automatic question answering technology to go further,realizing the transition from massive text retrieval to querying domain knowledge base.On the basis of understanding question precisely,question answering can achieve a better performance of query response to the upper concept ontology and structured knowledge base of instance level facts.Domain specific question answering focuses on the construction of domain ontology and knowledge base,precisely understanding of questions,semantic parsing and query answering on knowledge base.In this thesis,the design and implementation of a question answering system of the military information domain based on the Semantic Web technology is presented,where the main work and contributions are as follows:1.According to the domain data from different sources,the extraction of information,the generation of structured date and the construction of domain ontology as well as knowledge base are carried out.2.A set of templates of Chinese natural language question understanding is proposed.After processing natural language questions,the templates do pattern analysis of the question dependency tree,annotate nodes in detail and extract trunk structures in dependency tree,so that the templates are able to give out corresponding semi-structured intermediate representations to solve questions.3.For the Chinese natural language questions of complex factoid type with multi-adverbial constraints,a semantic model of question analysis is proposed,which can divide a question of complex factoid to multiple simple factoid triples.Based on these triples,a structured query of question triples' set with multi-constraints is generated for query answering on the knowledge base of structured data.Finally,the work above is integrated to a system called MiliQA,and experimental evaluation is carried out to validate its effectiveness.