Detecting Fake and True News by Applying Text Analysis and Deep Recurrent Neural Network

A.Prof.Dr.Raid Rafi Omar Al-Nima
Technical College of Engineering for
 Computer and Artificial Intelligence
 Mosul, Northern Technical University
Mosul, Iraq
عنوان البريد الإلكتروني هذا محمي من روبوتات السبام. يجب عليك تفعيل الجافاسكربت لرؤيته. 

Karam Sameer Qasim Qassab
Technical Institute / Mosul
Northern Technical University
Mosul, Iraq
عنوان البريد الإلكتروني هذا محمي من روبوتات السبام. يجب عليك تفعيل الجافاسكربت لرؤيته.

Imad Idan Abed Al-Khalaf
Technical College of
 Engineering for Computer and
 Artificial Intelligence
Mosul, Northern Technical University
Mosul, Iraq
عنوان البريد الإلكتروني هذا محمي من روبوتات السبام. يجب عليك تفعيل الجافاسكربت لرؤيته.

Keywords: Deep Recurrent Neural Network, Fake News, True News, Text Analysis

Abstract
Detecting fake news is a significant topic; it is valuable for warning people and protecting them from the consequences of such news. In this paper, a Deep Recurrent Neural Network (DRNN) is applied for detecting or recognizing fake and true news. Text data is first exploited and pre-processed. The pre-processing includes tokenizing, converting to lowercase, and erasing punctuation. Then, data is translated into sequences of values, which are utilized in the DRNN. The DRNN involves multiple layers: the sequence input layer, the word-embedding layer, the Long Short-Term Memory (LSTM) layer, the fully connected layer, the softmax layer, and the classification layer. A useful database from Kaggle named Fake News Detection (FND) is used; it has a huge amount of data. The obtained result achieved 99.77% accuracy, which is obviously very high