Web based keyword extraction on crowdfunding projects with novel features

Most of the existing research on contradiction focuses on detecting social media-based fake news or fake charity crowdfunding projects, but research on fake or infeasible crowdfunding projects is very limited. Therefore, we focus on how to gain an understanding of fake crowdfunding projects based on keyword extraction and selection. We used 12 features to extract keywords from web based information about crowdfunding projects. We evaluate this method with crowdfunding projects. Experiments showed that BERT with DT provide results of 69.33% in F-measure, which is better than other algorithms.