Sentiment Analysis Of Review Text With External Knowledge

In recent years,along with the rise of artificial intelligence,the development of natural language processing technology has received great attention from the industry,and sentiment analysis is one of its important research directions.The booming development of many social media platforms and e-commerce business has provided enough channels for users to express their comments and opinions.For example,users can post their opinions and views on current hot social events on social media platforms such as Weibo and Xiaohongshu anytime and anywhere.Customers can post their comments and experiences of goods and services on e-commerce platforms such as Jingdong and Meituan.These comment texts contain the emotional tendencies of the people and are of great research significance.Customers are able to decide whether goods are worth buying based on other consumers’ comments,enterprises can make improvements to the marketing of goods or services based on customers’ comments,and the government can grasp public opinion guidance and make corresponding policy adjustments in a timely manner based on the public’s remarks and opinions.Therefore,aspect-based sentiment classification based on social media comments has important research value.At present,aspect-based sentiment analysis based on deep learning has achieved significant results,but there are still some problems,such as: the comment text is often targeted,short and complex expressions,and it is still difficult for the machine to identify the sentiment tendency of specific aspects of the comment text in complex contexts;the cost of manual labeling of sample data is high,and the lack of training samples will limit the performance of the model.Therefore,this paper improves the model by combining syntactic knowledge and graph convolutional networks with external knowledge as supplementary information to improve the performance of aspect-based sentiment classification task.The main work of this paper is as follows:Some new evaluation objects and expressions often appear in comment texts,and in order to accurately identify the semantic information of aspectual words,this paper proposes a sentiment classification model based on dual-attention fusion knowledge.Firstly,the set of concepts related to aspect words is retrieved from the external knowledge base;then the knowledge features with higher degree of semantic relevance between the concept set and the context are weighted using two attention mechanisms;finally,the two parts of information,context and external knowledge based on aspect words,are combined as the final classification feature representation.The model’s understanding of the text depends not only on the understanding of lexical semantics,but also on the assistance of syntactic knowledge.In this paper,we propose a sentiment classification model based on the synergistic fusion of syntactic,semantic and external knowledge.First,the semantic features and syntactic features of the comment text are encoded using a bidirectional long and short term memory network and a graph convolutional network,respectively,and a multi-interaction attention mechanism is designed to obtain a better contextual representation from both semantic and syntactic perspectives;second,the contextual representation is enhanced by using the semantic relations of words in the external knowledge source Word Net,while the attention mechanism based on aspectual words focuses on weighting and aspectual word-related Finally,the contextual feature representation incorporating syntactic and external knowledge is used for sentiment classification by using a synergistic fusion mechanism of multiple features from local to global.The experimental results show that the inclusion of syntactic and external knowledge can effectively improve the classification performance of the model.