Research And Implement Of Question Answering System For Reading Comprehension

Question Answering System(QAs)is a promising research direction in the field of artificial intelligence and natural language processing.Question answering for documents is a complex research problem in natural language processing because of the rapid growth of articles and the diversity of text expressions in article.Text understanding and reasoning is one of the long-term goals for artificial intelligence research,particularly for reading comprehension.A model with the ability to reason about objects and their interactions is required for the development of question answering(QA)research.At present,the task of question answering is automatically answering a natural language question by understanding unstructured collection of natural language documents.For the best of our knowledge,none of the current question answering system can perform great when reasoning,especially multi-step reasoning.This topic uses the model based on memory network and the model based on relational network to solve the reasoning problem in the question answering system for reading comprehension.The core of the research work is to model attention model,which combines text features and question features to locate the answer object in the text to get the answer.Generally,the general model consists of three parts:text feature extraction module,multi-layer attention module and answer prediction module.The main work of this subject is as follows:(1)In the aspect of text feature representation,the corresponding text features are obtained from three aspects:letter level,word level and context level.Through multi-level feature extraction,text information is extracted from many aspects to improve the accuracy of experimental results.(2)Question Answering system realizes multi-step reasoning.Memory network and relational network model are used to achieve reasoning,and the attention mechanism in the model is improved,which significantly improves the effect of the model.(3)Make full use of the incremental information provided by keywords in the implementation of sentence-level question answering system.(4)On the basis of improving the model of memory-based network and relational network,a question answering system for reading comprehension task is designed and implemented.