The scale of false information present is vast, and this fake news has its consequences. A recent incident in Iran compelled the Indian Institute of Technology, Kanpur to find a solution to this problem. Many people in Iran lost their lives believing from false news that industrial alcohol could cure COVID-19 . Misinformation in times of a pandemic can be fatal and should definitely be avoided. Thus, Dr. Swaprava Nath took this project to make people aware of the fake news and stay safe.
Dr. Swaprava Nath, Department of Computer Science and Engineering, IIT Kanpur, and his team have developed a chatbot for contact tracing and fake news verification. Titled ‘Satyanweshi’ ('seeker of truth' in Sanskrit), the AI chatbot is programmed to find and check accurate information about COVID-19.
“Along with COVID-19, we have been infected with an ‘infodemic’, which is bombarding us with tonnes of fake news. To help you out, we introduce Satyanweshi, your AI assistant for truth checking,” IIT Kanpur’s press communiqué mentioned. Thus, showing their commitment to the problem.
How to use Satyanweshi?
A beta version of this solution is already available for public use. You can go to www.satyanweshi.net to access it or save the number +1-4155238886 as Satyanewashi in your contact list.
When a user receives the news on WhatsApp, which they want to fact-check, they can forward the message to the Fact Checker bot and get a result. The result clarifies whether the news is false or not. Apart from providing fake or real clarification, if the news is fake, it tells us why the news is fake.
Users can also clarify the information related to COVID-19 by just typing if they are unable to forward the news to the chatbot.
Satyanweshi has expanded to Messenger. Their beta version is now available on their website.
How does it work?
When any news is forwarded to the Satyanweshi contact, it applies an input processing technique. The news is then classified into pictures or text. Both of these are processed by Natural Language Processing techniques and tested against a model. The base training of the model comes from the trusted sources of the World Health Organization (WHO) and the Centre for Disease Control and Prevention (CDC).
There may be cases when manual intervention is required, but it happens when the data has been encountered for the first time by the database. In such cases, the data is clarified by a reliable source, and the result is sent to the user. This new data is learned by the model.
The model uses Binary Classification, that is, data can be separated into two groups; in this case, its real or fake. The machine learning model would generate a percentage of news that is false, and by Expectation-Maximization algorithm set a threshold above which the article will be fake. Below it, the article will be real.
Advantages of Satyanweshi
Satyanweshi is much more user friendly than most of the other fact check apps. A simple forwarding the message gives the result. Most of the fact-checking apps rely on copy-pasting the text from text or typing the text from a picture into the fact-checking app to provide the result. This gives it a wider reach, and with further implementation, it can be used to detect news from other languages. It can also be accessed in different social media applications like Twitter, Facebook Messenger, Instagram, and Telegram.
The most advantageous feature of the fact-checker bot is its high accuracy. Its accuracy is much better than existing fake news detectors. This is mainly because the existing fake news applications work for all kinds of news while the IIT Kanpur fact-checking bot’s model is trained only to detect COVID-19 related information. This model further improves by manual detection, thus increasing accuracy even more.
Satyanweshi has been recognized by the Indian Government and Indian Media. It secured 2nd place in MHRD AICTE SAMADHAN competition in response to COVID-19 among 2500 participants.
Source of the cover : 'Fact and Friction' from The Week