A study on a fake news identification model based on word embedding method

Status of information technology and mediaization Fake news is emerging as a permanent problem in our society. By developing a model to detect fake news, we aim to deliver reliable information about the impact of the current coverage of fake news. Among natural language processing methods, we would like to introduce various embedding methods to share fake news and share the performance of the model through a deep learning model based on the word embedding method. The push embedding method is a method of extracting meaningful features from news text data and identifying meaning and consistency between words. This method is used to identify information that does not match the actual content of the news article and place importance on fake news. After generating the embedding matrix of each word embedding method, TF-IDF, Word2Vec, and FastTextt, and combining the embedding layer with the deep learning-based LSTM model, which is a model with fake news, the power (accuracy) of the model is compared to see which is superior. An embedding method was presented. Comparing the cooperation of the models across participants in this study, we show that the Word2Vec method outperforms TF-IDF and FastText.