Design And Implementation Of Document Sentiment Classification System Based On Graph Neural Network

With the advent of the "big data" era,the number of texts containing user opinions on the Internet continues to increase,which gives many organizations or individuals opportunities to understand public opinion,sentiment,and social needs.However,at present,in many scenarios,it is still manual to analyze text comments containing opinions.This process not only takes time and effort,but also increases with the number and length of comments.Therefore,this article researches and implements a sentiment classification system based on graph neural network to help users quickly analyze the sentiment tendency of the text and extract key information.Document-level sentiment analysis refers to analyzing one or more articles and giving the sentiment tendency of each article.However,some existing sentiment analysis algorithms have a certain gap in performance compared with general topic-based classification algorithms.In addition,some existing keyword extraction algorithms have some limitations,such as extracting only through a given document collection,or extracting only for a single article,but the performance of the extraction results is not as good as the former.Therefore,to further solve practical problems through sentiment classification algorithms and keyword extraction algorithms,this paper studies and applies some sentiment classification algorithms and keyword extraction algorithms,and based on these algorithms,realizes a sentiment classification system based on graph neural network.The main research content includes the following three aspects:(1)First,the paper studies sentiment analysis algorithms and their applications based on Text-GCN and Fast-GCN models.For the previous machine learning and deep learning methods,either completely ignore the order of words,or only consider local or continuous word co-occurrence information.The Text-GCN model realizes global word co-occurrence information by constructing a text graph.Experimental results show that the Text-GCN algorithm has achieved better prediction results on some data sets.(2)Then,the paper studies the keyword extraction algorithm based on TopicRank and its application.Aiming at the problem that the commonly used statistical-based keyword extraction algorithm needs to input a document collection,or the performance of some TextRank-like algorithms is not as good as the former,this article considers using the TopicRank algorithm for keyword extraction.The extraction algorithm can not only meet the requirements of extracting keywords for only one article,but also can achieve performance comparable to extraction algorithms such as TF-IDF in terms of performance.Experimental results show that on some new data sets,TopicRank has achieved better prediction results than some commonly used keyword extraction algorithms.(3)Finally,this article uses Django,Python,Chart.js and other technologies to implement a sentiment classification system based on graph neural network.The system includes multiple modules such as Web services,log storage,and core algorithms.At the same time,this paper gives a detailed design and implementation method of the system in combination with the requirements.The final system test results show that the sentiment classification system based on graph neural network meets the demand and realizes the original vision.