Research On Precise Search And Visualization Of Online Social Network Objects Based On Semantic Matching

With the popularity of social networks,users can express their opinions on social media at any time and evaluate the content released by others,which has led to many social topics.Some users have great influence in social networks,and the information published can often cause people to pay attention and spread,and even change the public opinion of the topic,that is,the opinion leaders in the topic.Therefore,it is of great research significance to conduct precise search and sentiment analysis on topics opinion leaders to achieve real-time monitoring of topic development trends and stability of online social networks.The main contributions of this thesis include the following four aspects:(1)The realization of the acquisition and data preprocessing of online social network objects.Aiming at the problem that the data of social network objects is complex and difficult to obtain,an online social network object acquisition method is proposed.We use web crawler to obtain information on online social network objects,including user data and text data sent by users,and filter noise information that is widely present in social network data.Perform further field extraction and perform some basic preprocessing operations on the extracted data.Store the processed data in the database to complete data acquisition and preprocessing.(2)The realization of the semantic representation and matching of online social network objects.Aiming at the problem of semantic sparseness of short text data of social networks,a microblog semantic matching algorithm based on pre-training and Siamese networks(PSNSM)was proposed.For the text data sent by the social network objects,the text vectorization is carried out first.Further training and fine-tuning through the pre-trained model,fully exploit the correlation between text contexts,better grasp the semantic focus,and thus better model the text content to obtain a higher level of microblog-level semantic representation.The semantic matching training of microblog vectors is carried out through an improved Siamese network semantic matching model.We further optimizes the semantic representation of microblog and solved the problem of short text modeling,and realized the semantic representation and matching of microblog text.(3)The realization of the precise search and sentiment analysis of online social network objects.For online social network object search,it is mainly to find opinion leaders on specific topics.Aiming at the problem that most of the existing opinion leader search algorithms only focus on the relationship between user attributes and users,and ignore the microblog information sent by users,this thesis proposed an opinion leader search algorithm based on the double-graph model(DGRanK).Combining user information in social networks with microblog information sent by users,we adopt the correlation among microblogs to supplement the relationship between users and form a unified double-graph model.It enables precise search of opinion leaders and solve the problem that the existing opinion leader search algorithms does not make good use of all relevant data.A sentiment analysis algorithm based on pre-training and TextCNN(PTEAM)is proposed to conduct sentiment analysis on opinion leaders.The pre-trained model was used to model the semantics of microblog text to capture the semantically dependent information in the text.Then we use the improved TextCNN for further emotional focus.By making full use of the feature information of the whole text,we can effectively identify the emotional tendency of the text.The emotion of opinion leaders is obtained through the emotion of the text to realize the emotion analysis of social network objects.(4)The realization of the precise search and visualization system of online social network objects based on semantic matching.The system is divided into three modules:the data acquisition module implements the social network object data acquisition and data preprocessing;the object search module implements the semantic representation and matching of microblog text,and then complete the precise search of online social network objects based on the microblog semantic representation,and finally complete sentiment analysis of social network objects through sentiment analysis algorithms;the object visualization module systematically displays the user's social attribute data through visualization tools to complete the visualization of social network objects.This thesis implements the acquisition and data preprocessing of online social network objects,the semantic representation and matching of online social network objects,the precise search and sentiment analysis of online social network objects,and the precise search and visualization system of online social network objects based on semantic matching.It can realize precise search and sentiment analysis of opinion leaders,and monitor the development trend of topical events.