Identifying Topics: Analysis of Crowdfunding Comments in Scam Campaigns

摘要

In the Internet era, the effect of reviews and comments is reinforced as one can make a well-informed decision based on the experiences of others. Crowdfunding site, Kick-starter presents a platform to the backers to leave their feedback on a campaign. However, the comments are in abundance and diverse in nature that it becomes barely possible to wade through them to pick up the desired information. This study takes a step to identify the hidden themes in these comments to discover the different topics of discussion in scam campaigns and then these topics are compared with the topics identified in genuine campaigns. Topic models such as LDA (Latent Dirichlet Allocation) have been used in many areas. We have also used LDA to extract the dominant topics in a document. We evaluated this model on both scam and non-scam campaigns comments. We observed that the resulted topics in each category are distant from each other.