Fake News Detection Using Bert Semantic Similarity

The swift progress of the applications of NLP (Natural Language Processing) techniques particularly in the domain of semantic similarity are quite beneficial for the detection of 'fake news', that is, dishonest news stories that comes from the non-reputable sources. Therefore, to tackle this problem by building a model supported a count vectorizer (using word tallies) or a (Term Frequency Inverse Document Frequency) TFIDF matrix, (word tallies relative to however usually they're employed in different articles in your dataset) which will solely get you thus so far. However, these models don't think about the necessary qualities like word ordering and context. It's completely feasible that 2 articles which are similar in their word count are going to be fully totally different in their actual definition. The information science community has responded by taking actions against the matter and also took initiatives like there's a Kaggle competition referred to as because the “Fake News Challenge� and Facebook is using AI to filter faux news stories out of users' feeds. Combatting the faux news may be a classic text classification project with an uncomplicated proposition. So, it is doable for us to create a model that may differentiate between “Real “news and “Fake� news? thus, a planned work on collecting a dataset to train the model to understand context of sentences which can in turn predict faux and real news.