Research And Implementation Of Automatic Question Answering Algorithm Based On Text Comprehension

Machine reading comprehension is a very challenging task in the field of natural language processing.Its key goal is to make the machine have human-like reading ability,so that the machine algorithm model can understand the text content and correctly give the answers.The improvement of the under-standing of text content will be of key importance to the application fields such as information retrieval,dialogue system,and intelligent customer service.Its'development and promotion will change the way of human-computer interac-tion and achieve the real intelligence.Compared to choosing to fill reading comprehension tasks,segment ex-traction machine reading comprehension tasks are more challenging and rep-resentative,and currently are attracting more and more attention.This thesis fits into the current research trend,and also focuses on segment extraction ma-chine comprehension tasks to explore new technologies and new methods.The current large-scale datasets enables the construction of complex deep neural networks.This thesis will adopt the deep learning method to build model to the original passage and the question and predict the corresponding answers.The current text comprehension model mainly focuses on the semantic matching of the word dimension,and returns the part of the original text that has the greatest semantic relevance to the question as the answer.This kind of strategy can answer the question correctly under normal circumstances,but because the overall semantics of the question sentence is not considered,the model may not undeirstand the true questioning intention.and finally r-eturn the wrong answer.The original text with the most relevant content to the ques-tion is not the correct answer.Aiming at this problem,this thesis proposes a semantic summary mechanism for question statements and a semantic filtering mechanism for text,so that the model can understand the intent of the questions more accurately.The model proposed in this thesis has a significant improve-ment effect on the verification data set,the exact matching rate of the answer is increased by 1.1%,and the fuzzy matching rate of the answer is increased by 0.7%.In addition,based on the research of text comprehension automatic question answering algorithm,we implements an automatic question answering algorithm prototype system to showcase our research results.