Chinese Word Sense Disambiguation And Question Comprehension For Automatic QA

With the rapid development of artificial intelligence technology and the widespread application in daily life,people are seeking artificial intelligence in all walks of life.In customer service,human-computer interaction and other fields,automatic question and answer is an indispensable and urgent problems to be solved.Among them,the automatic question-answering for the open field is receiving more and more attention.In the au-tomatic question-answering question,we need to combine a variety of specific tasks and techniques of natural language processing.Therefore,the improvement of the effect of natural language processing tasks and the combination of these techniques are two impor-tant aspects of the research.This paper studies two natural language processing problems in the context of automatic question and answer,and combines these techniques to develop a prototype system of automatic question answering.First of all,we studied Chinese word sense disambiguation technology based on LSTM.Whether in Chinese or in foreign languages,the word polysemy is widespread.How to determine the semantic meaning of polysemy in the context is an urgent problem to be solved.In this paper,we propose a new word-sense disambiguation strategy based on synonym substitution,which transforms the problem into a text classification problem.What s more,a word-disambiguation data set is constructed based on this strategy,and a text classification model based on LSTM is used to solve the transformed classifica-tion problem.The experimental results show that the model converges successfully on the dataset,and the accuracy of the model is 78%,which achieves the correctness of the word-sense disambiguation method that trained a model for each word.Second we studied the questions understanding for automatic question answering.When conducting automated quizzes,understanding the user 's problem and understand-ing the user 's intent is a crucial step.Based on Hownet,this paper presents a new method to calculate semantic similarity based on synonym replacement strategy.Combining with word editing distance and word vector cosine,this paper proposes a measure method of syntactic similarity of words.Applying this measure method to the property matching problem of automatic question answering,the whole algorithm of automatic question an-swering is completed by combining the entity linking technology.The experimental re-sults show that the F1 value of automatic question answering algorithm after applying this metric algorithm reaches 0.83,confirming the feasibility and effectiveness of the compre-hensive similarity measure method.Finally,we built an automated question answering system based on WeChat public platform.Although the automatic question answering system is open to the field its func-tions are always limited.In this paper,the scope of the function of automatic question and answer is determined from the reality,an automatic question answering system prototype is designed and implemented and it is deployed on the WeChat public platform to realize the automatic quiz of WeChat public account.After testing,this system completes all the functions of automatic question answering system design well.