‘Tanbih’ - Our Answer to Fake News

Issues surrounding fake news have been on the rise in recent years, and for people who rely on online platforms for their daily dose of news, the question remains: how do we know that what we’re reading is valid?

Qatar Computing Research Institute (QCRI) might just have an answer to that! QCRI, a research institute under Hamad Bin Khalifa University, is dedicated to performing research in areas of social computing, cyber security, data analytics, and much more.

For the past four months, QCRI has been working on a news aggregator, ‘Tanbih’, that will fact-check the news before it has even been posted! The purpose of this is to ensure people are aware of what they’re reading.

Here are four things you need to know about Tanbih:

1. What makes Tanbih different from other fake news algorithms?

The idea of an algorithm that detects fake news might not initially sound like something new - but Tanbih actually is! Most current anti-fake news efforts focus on checking claims, rumors, or news articles, which can be very costly. But to make it more time- and cost efficient, QCRI’s aggregator actually checks the news outlet itself, rather than each individual article - and it can also check news outlets in different languages! The model has been applied to Arabic, Hungarian, Italian, Turkish, Greek, French, German, Danish and Macedonian! And while the English model is most accurate, a language-independent one isn’t far off.

2. How does it actually work?

Tanbih initially estimates the news outlets’ factuality by looking at different factors: articles, social media and the website itself!

The aggregator will study several hundred articles published by a news outlet and analyze the style, subjectivity, sentiment and vocabulary richness, to name a few criteria. It will also analyze the outlet’s Wikipedia page for bias. Other factors that play a role are whether the news outlet is present on social media - and if so, whether the account is verified, how old it is, the number of followers it has, and the different links on the profile. Finally, it checks the website’s URL structure, and the web traffic to the website to determine its popularity.

Based on this, it creates a profile for most major news outlets that determines whether they are likely to publish fake news or not. It also determines their overriding political ideology (left vs. right), as well, hyper-partisanship, and level of propaganda.

Preslav Nakov[10]

Dr. Preslav I. Nakov is leading QCRI’s work on Tanbih

3. But is it successful?

Success rates currently vary, depending on the content. As of now, Tanbih is 65% accurate at detecting whether the news outlet has a high, medium, or low level of factuality, and a 75% success rate for detecting if it’s politically right-, left- or center-leaning.

4. So what’s next?

Tanbih can’t be used publicly just yet. Currently, it is a research prototype that’s still being developed by QCRI’s team. The ultimate goal is to create a news aggregator that automatically and accurately predicts the factuality of news reporting, its bias, and the likelihood of the article being propagandistic, among other features.

This isn’t the only cool project QCRI has worked on. There are a number of other innovative ideas the institute has conceptualized, such as the Qatar Center for Artificial Intelligence, which focuses on leading AI research in Qatar and the region! If you want to know more about QCRI, check out their website here.