Detecting Fakeness in Urdu News using Web based application

Fake news is created and published with the purpose of deceiving readers in order to profit politically, and it frequently uses sensationalist, exaggerated, or obviously untrue headlines to do so (Bachenko, Fitzpatrick, & Schonwetter, 2008). Because of their widespread

Project Title

Detecting Fakeness in Urdu News using Web based application

Project Area of Specialization

Computer Science

Project Summary

Fake news is created and published with the purpose of deceiving readers in order to profit politically, and it frequently uses sensationalist, exaggerated, or obviously untrue headlines to do so (Bachenko, Fitzpatrick, & Schonwetter, 2008). Because of their widespread dissemination in public discourse, particularly on social media, and their worrying impact on our lives, fake news has become a problem of crucial importance in our society (Lazer, et al., 2018). Fact-checking websites, are typically limited to a single topic of interest, such as politics, and require human knowledge, making it difficult to gather datasets with some degree of generalization over several domains (Pérez-Rosas, Kleinberg, Lefevre, & Mihalcea, 2017)

As Urdu is a low resource language, We are making our corpora of about nine thousand Urdu news of different categories which includes Political, Business, Celebrity, International etc ,including 6,000 True news from official websites of Urdu news channels  which includes ARY Digital, 92 News, Hum News, Nawaye Waqt News, Dawn News, Express News, Samaa Tv, 24 News, Bol News, Ptv news and 3,000 Fake news from twitter by using python scraping script. 

For model training part, Preprocessing is done in which we have used TFidVectorizer. Then, Feature Extraction is done through Bidirectional Encoder Representations from Transformers (BERT) and Robustly Optimized BERT Pre-training Approach (RobertA) . The next step is of training the model using different deep learning classification techniques which includes Passive Aggressive Classifier, Recurrent Neural Networks (RNN), Long Short Term Memory Network (LSTM) which will predict the news as if it is Fake or Real. The news entered by the user will be stored in the database at the backend. The user will then get the result of the prediction on the screen.

Project Objectives

1. To propose an approach for automated classification of news as fake or real using deep learning algorithms.
2. To exploit news content which predict whether the news is fake or real.
3. To explore different contextual properties that can be used to distinguish fake news from real.
4. To make the status of Urdu language equal to other rich resource languages like English, Chinese, Spanish, and Portuguese by adding to its resources.
5. To achieve high accuracy of trained model by using various deep learning algorithms/classifiers/techniques.

Project Implementation Method

Our trained model is deployed on a web based application which mainly focusses on detecting fake news in Urdu language. We have added different pages in our website which includes Home Page, Prediction Page and a Page for including Research Papers on the header of the website. A direct link to Prediction Page is also added to Home Page. 

Benefits of the Project

Today, fake news has become a serious issue that is spreading havoc around the globe. As a result, developing an algorithm that is as accurate as possible will be a revelation, and it will have a huge impact on present societal concerns as well as the current political situation. People use social media and online news articles as a primary source of news and data because they are easy to access, have a low cost, and are readily available—just a click away. However, it has a number of drawbacks, including the lack of a check on the source, as well as the legitimacy and validity of the viewpoints being endorsed. As a result, we've suggested a novel technique for detecting false news that includes contextual relationship as a key element to boost accuracy.

Technical Details of Final Deliverable

1. SSD
2. For detecting fake speech, a microphone and good recording system is needed
3. For identifying fake images, we need high quality camera and scanner
4. Braille Keyboard for blind

Final Deliverable of the Project

Software System

Core Industry

IT

Other Industries

Media

Core Technology

Big Data

Other Technologies

Sustainable Development Goals

Decent Work and Economic Growth, Peace and Justice Strong Institutions

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com