Leveraging Commonsense Knowledge on Classifying False News and Determining Checkworthiness of Claims

Widespread and rapid dissemination of false news has made fact-checking an\nindispensable requirement. Given its time-consuming and labor-intensive nature,\nthe task calls for an automated support to meet the demand. In this paper, we\npropose to leverage commonsense knowledge for the tasks of false news\nclassification and check-worthy claim detection. Arguing that commonsense\nknowledge is a factor in human believability, we fine-tune the BERT language\nmodel with a commonsense question answering task and the aforementioned tasks\nin a multi-task learning environment. For predicting fine-grained false news\ntypes, we compare the proposed fine-tuned model's performance with the false\nnews classification models on a public dataset as well as a newly collected\ndataset. We compare the model's performance with the single-task BERT model and\na state-of-the-art check-worthy claim detection tool to evaluate the\ncheck-worthy claim detection. Our experimental analysis demonstrates that\ncommonsense knowledge can improve performance in both tasks.\n