Research And Application Of Chinese Question Generation Method Based On Deep Learning

With the rapid development of NLP and knowledge graphs,intelligent Question Answering(QA)has become an important way to quickly acquire and learn knowledge in professional fields.Among them,question generation,as an important subtask in intelligent question answering,has attracted more and more attention.However,most of the existing research on question generation focuses on English corpora,and there are few researches on Chinese.Especially for Chinese paragraph texts,most of the existing methods only generate a single question,lacking a complete understanding and summary of the key knowledge in the entire text.In order to fully analyze each sentence in the paragraph,this paper uses a deep learning model to extract multiple core sentences containing knowledge points from the paragraph text,and then generate related questions for each core sentence.By combining mechanisms such as Chinese sentence-level tag embedding and word case embedding representation,the semantic acquisition capabilities of both the extraction and generation models have been improved,thereby achieving more complete paragraph information retrieval.The research content and innovation points of this paper mainly include the following aspects:(1)In order to make full use of the context information of the sentences in the paragraph text,this paper proposes the GCSS Chinese core sentence extraction model.This model combines Seq2Seq and CNN models,and applies the Sentence-Embedding mechanism to add the length and boundary information of each sentence in the paragraph text to the word embedding layer to improve the extraction effect of the model.Compared with the traditional single data classification model,the model proposed in this paper has improved Recall and F1-score on the core sentence extraction data set.(2)In order to make full use of Chinese lemma information and location information,this paper proposes the CQGLM-PE Chinese question generation model.The model incorporates the PLattice mechanism to integrate the lexical information and relative position information of the Chinese input corpus into the model,and combines the ECopy mechanism to improve the correlation between the input and output of the model.A large number of comparative experiments show that the CQGLM-PE model has achieved better performance in the question generation dataset.(3)Based on the proposed core sentence extraction and question generation model,this paper designs and implements an intelligent question generation system.The test results of the system show that the question generation model proposed in this paper can be applied to the actual system,meet the actual needs of natural language automatic question generation,and realize high-quality question generation.