Research And Implementation Of Domain Intelligent Question Answering System Based On Semantic Similarity

The rapid development of internet technology has led to a surge in data volume in the network.How to efficiently help users obtain specific domain information from the massive data on the internet is an important issue,and intelligent question answering systems are a good way to solve this problem.Intelligent question and answer system refers to a technology that accurately answers users’ questions in the form of one question and one answer,and provides personalized information services to users through interaction with them.Semantic Similarity is a technology to study the relationship between different texts by calculating the similar information of two texts.It is an important research direction in the field of natural language processing.Especially in the information exchange of intelligent question answering systems,there are advantages in efficiency and accuracy.At present,in intelligent question answering systems,especially in specific fields,it is necessary to develop a domain intelligent question answering system based on semantic similarity,as traditional methods mostly rely on relationships for question answering and have insufficient efficiency in preprocessing knowledge bases and questions.For the key triplet sorting problem in the intelligent question and answer process,this thesis uses semantic similarity method to solve the problem,and divides triplet sorting into two tasks: coarse-grained triplet sorting and fine-grained attribute triplet sorting.For coarse-grained triplet sorting,this thesis proposes a Text Semantic Matching Model based on Attention Inter Networks(TSM-AIN)for triplet sorting.Firstly,the model integrates the matching extraction of representational networks with the interaction matrix in interactive networks;Secondly,the mixed common attention mechanism is used to receive complex feature information and strengthen the feature information of important texts in text pairs,so as to realize the information exchange between text pairs;Then,for the high-dimensional features output by the hybrid common attention module,the key feature representation between text pairs is processed through the cyclic network of the dynamic attention matching matrix,and the key features are saved.For fine-grained attribute triplet sorting,this thesis proposes an Attribute Matching Algorithm based on Mixed Similarity Calculation(AM-MSC)to perform the final sorting on the results of coarse-grained triplet sorting.Finally,the feasibility and effectiveness of the proposed model and algorithm were verified through comparative experiments with other models and algorithms.Based on the key technologies mentioned above,this thesis designs and develops a domain intelligent question answering system based on semantic similarity.This system provides two types of Q&A methods: general domain and specific domain(such as finance),with main functions including data integration,knowledge base construction,intelligent Q&A,and user management.At the same time,this thesis adopts the B/S architecture to build the system,and is developed using Python language.Finally,system testing shows that the system constructed in this thesis has implemented all requirements and provided intelligent Q&A functions for both general and specific domains(such as finance).