Research On Knowledge Graph Representation And Reasoning For Multi-hop Question Answerin

Nowadays,the Internet era is full of huge amount of information,and it is difficult for people to retrieve and obtain the required target knowledge efficiently.To address this need,Multi-hop KGQA uses knowledge graph as knowledge source to obtain answers to complex questions posed by users through multi-hop reasoning.Although the existing work has achieved a series of results,it still faces many challenges in practical applications.On the one hand,knowledge graph faces the problems of low computational efficiency and data sparsity in the process of use,which affect the subsequent reasoning task.On the other hand,Multi-hop KGQA should have the ability to handle complex problems and provide explainable processes,while existing work often fails to take into account these two points.To this end,this paper proposes a knowledge graph embedding model based on k-order sampling and graph attention networks and an efficient multi-hop reasoning network for question answering based on topological subgraph and attention redistribution mechanism,and also applies the proposed model algorithm to an intelligent question answering system for SDH optical transmission based on knowledge graph.The specific work and contributions are as follows:(1)A knowledge graph embedding model based on k-order sampling algorithm and graph attention networks.Knowledge graph embedding(KGE),which aims to map entities and relations of knowledge graph into a low-dimensional space to obtain its vector representation.Existing KGE models only consider the first-order neighbors,which influence the accuracy of reasoning and prediction tasks in knowledge graph.In order to solve this problem,a novel KGE model based on k-order sampling algorithm and graph attention networks is proposed.Firstly,a k order sampling algorithm is proposed to obtain the neighbors’ features of a central entity by aggregating k-order neighborhood in the pruned subgraph.Then,the graph attention networks are introduced to learn the attention values of the central entity’s neighbors,and the new entity embedding is obtained by the weighted sum of neighbors’ features.Finally,the Conv KB is used as a decoder to analyze the global embedding property of a triple.Evaluation experiments on several datasets,WN18 RR,FB15k-237,NELL-995,Kinship,reveal that the model performs better than the state of-the-art models on the task of link prediction.Besides,the influence on the model hit rate while changing order k or sampling coefficient b has been discussed.(2)An efficient multi-hop reasoning network for question answering based on topological subgraph and attention redistribution mechanism.AS a multi-hop knowledge graph question answering model,Transfer Net performs reasoning at each step by multiplying the entity score vector and the relation score matrix,which means all entities of the knowledge graph are taken into the calculation.Although its reasoning is efficient and transparent,but its reasoning on large-scale knowledge graph leads to relatively high time complexity.In order to solve this problem,we propose an efficient multi-hop reasoning network for question answering based on topological subgraph and attention redistribution mechanism.Specifically,a path expansion algorithm is proposed to form a topological subgraph of the question,thus reducing the computational complexity of reasoning.In order to increase the reasonability,we design an attention redistribution mechanism to reduce the weight of historical reasoning information,thus enabling each step to mainly focus on the current reasoning part of question.Experiments on four datasets,Meta QA,Meta QA Text,Web QSP and Comp Web Q,show that the accuracy of our model is close to Transfer Net,but the time efficiency significantly better than the latter.(3)Finally,based on the above model algorithm,an SDH optical transmission intelligent question answering system based on knowledge graph is built in the field of SDH optical transmission of National Grid Communications.The system is able to reason out the correct answer in the knowledge graph by voice recognition of the question spoken by the user,and then return to the user in the form of voice.The system helps users to obtain SDH knowledge quickly and accurately,reduces the pressure of manual customer service,and has certain practicality.