Using Explainable Artificial Intelligence and Knowledge Graph to Explain Sentiment Analysis of COVID-19 Post on the Twitter

Social media has become the common way for people to share information and opinions. For analyze opinions about events or products, sentiment analysis has become a hot topic. The existing methods mainly use deep learning models to directly analyze the data collected on the social media and get results. However, deep learning models are generally “black boxes”, and it is impossible to know the relationship between data and results and the adjustment of internal parameters of the model. Therefore, the concept of explainable AI has become important. Explain the results of deep learning models through explainable AI models. In addition, knowledge graphs are constructed through reliable external databases. A reliable relationship diagram can also provide a good explanation effect. The study uses explainable AI and knowledge graphs to assist in explaining the application of deep learning models in sentiment analysis to understand which features in the data are important features, as well as the influence and attributes of these features. In the results, explainable AI specifically shows the impact of each feature in the sentence on the result, and the knowledge graph uses the sentiment of keywords, the sentiment of the post itself and the Subject–Verb–Object knowledge graph formed by the common verbs in the post to assist in explaining the results of the deep learning model. Finally, the concept of explainable AI and knowledge graph can help users better understand how features affect model detection.