Sentiment Summarization Generation Based on Multi Instance Learning and Graph Convolution on Social Media

Sentiment summary generation task is very valuable in social media, which can extract different aspects from people’s evaluation of products or services, such as the quality of products, and generate summarized opinionated sentiment summaries. And nowadays, sentiment summarization still suffers from insufficient contextual information capture and incorrect and comprehensive aspect extraction. In this paper, we propose a new strategy to solve the above problems: firstly, MILNET is used to automatically identify aspects related to a product or service from a large number of user reviews, and then the results are inputted into GCN, which is used to enhance the importance of contextual words close to the aspect; meanwhile, E-MABA is utilized for further optimization of the model on aspect extraction. By conducting a large number of experiments on the OPOSUM dataset and comparing the existing state-of-the-art methods, it is found that our method achieves better performance than other methods, proving the feasibility and effectiveness of our proposed strategy.