Sentiment Knowledge Graph: Explore and Evaluate the Social Media Content

Sentiment analysis on social media has gained popularity because of its impact on the broad spectrum of applications like recommendation systems. The facts derived in the sentiment analysis can be reused in scenarios with different structures depending on the needs, as in the case of information searching. This paper suggests a hybrid method for sentiment analysis and its representation in graph structure so as to make it easier for further analysis and refinement. The BERT model is used to find the sentiment value of the comments in the COVID-19 related Twitter dataset. A sentimental knowledge graph is created using the sentiment value and topics from the respective Bert and LDA topic models, along with the polarity of the tweets. The knowledge graph (KG) created for the dataset is effectively used in the exercises to derive the secondary facts in the experiments. It is observed that the effectiveness of the KG can be further improved by incorporating other sensible parameters as per the requirements of the problem.