Detecting Fake News Using Stance-Based Approach
People have become more aware about the problems caused by fake news these days. In the past decade, many researches have been conducted on how to spot fake news, and one of the methods is to use stance detection algorithm to obtain the relationship between truthfulness and proposition. In this paper, dataset from the Fake News Challenge will be used as the base for such stance detection algorithm. Once an article enters the algorithm, the algorithm would first search relevant articles from reliable news media, and then compare the propositions (agree, disagree, or discuss) between input article and its relevant articles. The results will be collected as a table and then apply machine learning techniques to determine the relationship between truthfulness and proposition. Word2vec and wordcounts will be used as medium to transform words into numeric features for analysis. The final result of this paper in section 5. showed that there is a significant relationship between truthfulness and percentage of stance for news articles. In order to use stance-based algorithm, a dataset consisting of 11,110 instance is manually labelled for stance based on an ongoing competition on Kaggle. Overall, the paper has demonstrated a new possibility for fake news detection.
Information science|Computer science
Lo, Su, "Detecting Fake News Using Stance-Based Approach" (2018). ETD Collection for Fordham University. AAI10932243.