A Hybrid Methodology (NN + LSTM) to Detect Fake News
Fake news, deliberate disinformation, sham news, parodies, satire, and hoaxes are present even before the invention of the Internet. But due to the arrival of the Internet, misinformation of the story has become easier. The globally accepted definition of pseudo-news is: “any fictitious articles/stories that are designed to misguide the readers” or “Any information that can’t be validated, with no sources too, and mostly false”. Hoaxes is just like beating around the bush. The main intention of publishing or releasing junk news is a part of psychological warfare or to increase the readership. In general, the goal of publishing fake news is getting profits through click baits. These click baits grab curiosity of the users besides lure them with snazzy headlines. They are designed to increase the advertisements to their products and to attract the users. This exposes emergence of social networking sites in the light of fake news.
This paper comes up with the idea to authenticate the news i.e., to detect whether the given statement of news is real or fake. This operation is done with the help of NLP and Deep Neural Networks.
The dataset used in the project consists of both fake and real NEWS headlines. After Data Preprocessing of the dataset, the model is trained by using Machine learning algorithms. This paper comes up with the applications of NLP (Natural Language Processing) techniques and is based on a countvectorizer and tfidf matrix and Deep Neural Network using either Tensorflow or Keras. On analyzing the results, we come up with the idea to combine both LSTM & NN to achieve fruitful results.
Keywords: Countvectorizer, LSTM, NLP, NN, tfidf.