A Recommendation Method for Social Media Users based on a Sentiment Analysis Model
Ying HongDa, Kosuke Takano · 2022 IEEE 4th Global Conference on Life Sciences and Technologies (LifeTech) · 2022
Sentiment analysis for social media users analyzes text and images that the users submitted to social media. However, it has been difficult for conventional sentiment analysis methods to properly analyze the sentiment of social media users in the cases that there is no relation between text and images that the user submitted, or there is the contradiction of the user's sentiment extracted from the text and images even if there is correlation between them. In this study, in the context of content recommendation according to user's sentiment, we propose a method for calculating user's sentiment score for text and images based on the correlation between the text and images and the difference of sentiment scores calculated for the text and images, respectively. In the experiment, we verify the feasibility of our proposed method by the evaluation using content analysis models and sentiment analysis models trained with a text dataset and an image data set.