Design And Implementation Of Chinese Question Answering System For Complicated Statements

Existing question answering systems cover template methods,graph query methods,presentation learning methods,and deep learning methods,but these methods lack the semantic understanding of Complicated Chinese question statements.The complexity of Complicated Chinese questions has the diversity of question types,the polymorphism of entities in question,and the ambiguity of question semantics,which lead to the poor performance of the question answering system in Complicated Chinese questions.In addition,The Question Answering System lacks a mature deployment plan,and the redevelopment caused by manual annotation,answer processing,system division structure,and demand changing affects the cost,reasoning ability,robustness,and iterability of the system.In order to solve the above problems,this paper proposes the design and implementation of a Chinese Question Answering System for Complicated Statements,based on the classification model to deal with the diversity of Chinese question types,based on named entity recognition and named entity disambiguation to deal with the polymorphism of question entities,based on the knowledge graph deal with the ambiguity of question semantics.The QA system is constructed based on solving the complicated statements.The paper designs and implements a Chinese question answering system for complicated statements and its deployment plan.The main research points and contributions of mine are as follows:(1)This paper designs deep learning models to solve the complexity of three aspects of Chinese Complicated Statements.The question classification model,is built based on the Bi-directional Long Short Term Memory networks,question types of Chinese complicated statements are determined.The paper proposes to extract entities of Chinese complicated statements based on named entity recognition technology,and use named dictionary to generate candidate entities,and builds the entity disambiguation model based on the Siamese network to eliminate the ambiguity of entities and generate the only entity of Chinese questions.This paper constructs a knowledge graph for the Chinese complicated statements.The relational path of the the Chinese complicated statements is generated,and the answers to the Chinese questions are generated,based on the types and entitys of the Chinese complicated statements.(2)This paper implements a prototype system of a question answering system for Chinese Complicated Statements,CCS-QA.At the data layer,this paper implements an entity recognition and entity disambiguation annotation strategy to expand and update the data of CCS-QA;at the model function layer,this paper implements a Chinese question comprehension module to query the knowledge graph,map the relational path of the Chinese question,and generate answers;at the interface layer,this paper implements RESTful API interfaces to provide third-party users with Chinese question answering services;At the application layer,this paper implements the deployment of a question answering system for Chinese Complicated Statements.(3)This paper evaluates the overall system performance through prototype system and experimental verification.In terms of classification of Chinese Complicated Statements,CCS-QA has reached an F1 value of 97.41%;in terms of entity recognition of Chinese question sentences,CCS-QA has reached an accuracy rate of 98.33%;in terms of entity disambiguation,CCS-QA has reached an F1 value of 95.57%;In terms of overall system performance,the prototype system proposed in this paper has reached an accuracy rate of 83.65% in the data set NLPCC-2016.The experimental results show that CCS-QA is feasible and effective.