Unlike the traditional fake or real news binary classification task, the fake news classification task under the Public Editor Group’s definition is a multiclass classification task. On the first layer, a text from a news article can have potential problems from Language, Evidence, Reasoning, Sourcing, and Probability perspectives. On the additional layer, for example, under the language category, a sentence or text can be labeled into eight subcategories: metaphor unhelpful, writing errors, exaggeration, inappropriately emotional, manufactured scandal, problematic tone, inappropriate slang, and denigration. 

Project Report

 

Term
Spring 2022
Topic
Social Sciences
Technical Area(s)
Natural language processing (NLP)