Research On Q & A System Based On Deep Logic Knowledge Extraction

Question Answering (QA) is a fast-growing research area which combines research from Information Retrieval, Information Extraction and Natural Language Processing. It is not only an interesting and challenging application, but also the techniques and methods developed from question answering inspire new ideas in many closely related areas such as document retrieval, time and named-entity expression recognition, etc.With the development of Internet it becomes the largest knowledge source. Many question answering systems solve some questions quite well using the explicit knowledge from big data. However much more knowledge is implicit in texts and a question answering system must make an inference to find right answers. However it is faced with challenges to extract implicit knowledge. Because web texts are nearly all unlabeled free texts. Moreover these texts are incomplete and noisy and contain many errors. So the traditional predicate logic inference is not qualified for knowledge from Internet.In this paper, we propose an approach to discover logical knowledge for deep question answering, which automatically extracts knowledge in an unsupervised, domain-independent manner from background texts and reasons out implicit answers for questions. The main contents of this paper are as following:(1) Firstly, natural language expressions are transformed into predicates in first-order logic using semantic role labeling which labels predicates and arguments more precisely. Semantic role labeling can identify more than two arguments and provide position information for disambiguation. So extracted relations are more precise and general.(2) Secondly, association analysis and statistical relevance are used to uncover the implicit relationship among predicates and build propositions with weights for inference. This method automatically constructs a general knowledge base.(3) Finally, Markov logic network is adopted as the inference system in order to overcome noise, indeterminacy and incompleteness of knowledge base. Experiments show that these propositions can improve the performance of question answering system significantly.