Research And Application Of Text Sentiment Analysis Based On Graph Convolutional Network

In recent years,artificial intelligence technology has enabled robots to become an indispensable aspect of human existence.However,there are some limitations to the interaction process between humans and robots,particularly in the processing of emotions.In order for robots to be able to emotionally comprehend and respond to human language,text sentiment analysis is a crucial technical means.This technique provides robots with the potential to achieve more human-like and natural interactions with humans.Traditional methods for text sentiment analysis have limitations in processing graph-structured data,particularly in dealing with relationship information between nodes.This requires manual design and can lead to problems such as sparsity and variability.In contrast,text sentiment analysis based on graph convolutional neural networks has a significant advantage because it can directly perform convolution operations on graph structures,adaptively capture node relationships,and extract feature representations.However,previous studies have overlooked the importance of syntactic relations in texts and lacked professional sentiment knowledge when dealing with texts from different domains.To address these issues,this paper proposes a new graph convolutional network method that can handle text sentiment analysis tasks in various fields through syntactic relationship interaction,external sentiment knowledge,and dependency relationship enhancement.To demonstrate the efficacy of the suggested methods,we applied it to an actual public opinion system and obtained positive results.The following are the paper’s primary contributions and innovations:(1)The existing research ignores the type of dependency relationship between words,which makes the model unable to make full use of the syntactic structure information in the text.In this paper,a fine-grained sentiment analysis method is adopted,and syntactic relation information and attention mechanism are introduced to better identify and utilize concrete entities and syntactic dependency in sentences,thereby achieving more accurate sentiment analysis.(2)Aspect-based sentiment analysis methods lack external affective knowledge guidance and may result in model distortion due to equal treatment of dependency relationships.This paper introduces an external affective knowledge base and a more high-frequency dependency method to enhance the representation of text characteristics and the performance of the sentiment analysis model(3)In this paper,the proposed methods are applied to the public opinion detection and analysis system,which proves that it can effectively improve the performance of sentiment analysis model,and provides valuable guidance for practical applications.