Due to the wide accessibility of the internet, people seek and consume information via social media because of its low cost, ease of access, and rapid transmission of information. However, the cost becomes expensive and dangerous when non-experts say anything. This research project is part of the problem of verifying the reliability of fake news. The goal is to propose a model that will allow readers to verify the veracity of the news they read. Our work involves finding a classification method based on several approaches to properly classify real and fake news. For this purpose, we have developed a two-phase solution, the first of which consists in classifying the veracity of the news (false or true), and the second phase deals with the verification of the reliability of the source of information. To do this, we applied natural language processing techniques. We evaluated the performance of our LSTM and GRU models using a precision, recall and F1-score metric on the prediction of fake news verification. The results show that the singlelayer LSTM and GRU models have high prediction accuracy (0.988 and 0.987, respectively), and that stacking layers does not have a significant impact on prediction accuracy. The two-layer models performed almost identically to the single-layer models. Finally, we were able to build a prototype.
Readership Map
Content Distribution
Due to the wide accessibility of the internet, people seek and consume information via social media because of its low cost, ease of access, and rapid transmission of information. However, the cost becomes expensive and dangerous when non-experts say anything. This research project is part of the problem of verifying the reliability of fake news. The goal is to propose a model that will allow readers to verify the veracity of the news they read. Our work involves finding a classification method based on several approaches to properly classify real and fake news. For this purpose, we have developed a two-phase solution, the first of which consists in classifying the veracity of the news (false or true), and the second phase deals with the verification of the reliability of the source of information. To do this, we applied natural language processing techniques. We evaluated the performance of our LSTM and GRU models using a precision, recall and F1-score metric on the prediction of fake news verification. The results show that the singlelayer LSTM and GRU models have high prediction accuracy (0.988 and 0.987, respectively), and that stacking layers does not have a significant impact on prediction accuracy. The two-layer models performed almost identically to the single-layer models. Finally, we were able to build a prototype.
