Identification of Misinformation Using Word Embedding Technique Word2Vec, Machine Learning and Deep Learning Models

EasyChair Preprint 11811

Abstract

Real-time news is widely disseminated through the internet on a global scale. One of the factors contributing to its success is the simple and speedy spread of news. Social networking platforms have a huge user base that includes people of all ages, genders, and social backgrounds. Considering these positive aspects, a serious drawback is the propagation of misinformation, as most individuals read and spread information without giving any thought to its veracity. Researching techniques for news authenticity is so essential. To address this problem, a fake news identification system is created by training the COVID-19 tweets with roughly 12427 records taken from Kaggle and GitHub repository from five different sets, annotated manually as Fake (0) and Real (1) by cross-checking through websites that verify facts using machine learning classifiers like RF, SVM, LR, NB, and Deep Learning classifiers LSTM and Bi-LSTM. The feature extraction process makes use of the Word2Vec word embedding technique. According to the findings, Bi-LSTM performed better than all the other models in terms of accuracy, scoring 87.3%.

Keyphrases: COVID-19 Tweets, DeepLearning, MachineLearning, NLP, Word2vec embedding technique, fake news, text classification

@booklet{EasyChair:11811,
  author    = {Arati Chabukswar and P Deepa Shenoy and K R Venugopal},
  title     = {Identification of Misinformation Using Word Embedding Technique Word2Vec, Machine Learning and Deep Learning Models},
  howpublished = {EasyChair Preprint 11811},
  year      = {EasyChair, 2024}}