The Construction And Realization Of People Social Relation Network

With the development of Internet,network reflect a contains a huge variety of complex information of unstructured text data,which implies many social connections between characters,extract the characters relations and realize social network visualization characters,to improve the quality of the related work from personnel of course of business and work efficiency is very important.By constructing the data set of character relation,this paper realizes the extraction of character relation,and builds the extraction and information inquiry system of character social relation.The main work of this paper is as follows:(1)At present,there are few researches on the task of character relationship extraction,and available corpus is limited,and manual annotation is time-consuming and labor-consuming.Therefore,this paper builds a certain scale data set of character relationship extraction by crawling the character relationship triad and the text content of the character entry on the interactive encyclopedia and using remote supervision technology.Aiming at the mislabeled data caused by remote supervision,this paper calculates the TF-IDF value to filter the noise data,and verifies the effectiveness of denoising in the experiment.Finally,this paper constructs 14,059 data extracted from character relationships,and carries out follow-up experimental research on 12 kinds of character social relationships.(2)Because the relationship extraction model is domain limited and there are few researches on the Chinese character relationship extraction field at present,this paper proposes a character relationship extraction model based on two-way GRU neural network.In order to obtain the semantic features of the sentence,the Bert pre-training model was used to generate the dynamic word vector representation,and the bidirectional GRU neural network was used to effectively learn the sentence context features.Aiming at the problem that irrelevant information in long text interfered with model learning,the attention mechanism was used to dynamically reduce the influence of irrelevant information on model performance.The comparative experiments show that the model is effective in both the present dataset and the public dataset.In addition,it is found that the size of training data of different relationship types and the semantic complexity of sentences also have a certain influence on the learning ability of the model.(3)In order to better apply the task of character relationship extraction to the actual scene,improve the utilization rate and display effect of triples,and provide more efficient tool of character relationship extraction and query for relevant workers,this paper built a character social relationship extraction and information query system.The Neo4 j graph database can be used to store,manage and visualize data more efficiently.The corresponding character network can be obtained according to the text sentence input by the user.Through the extraction of characters' social relations and the information inquiry system,the employees with special work needs can use the system to quickly clarify the social relations between characters and grasp the social details of characters.This paper studies the construction of the social relations network of characters,and realizes the extraction and practical application of the structured information of the social relations of characters.On this basis,it can further explore the reasoning and question answering system of the social relations of characters,which is of great significance for the in-depth exploration of the potential connections between characters.