4
$\begingroup$

I would like to work on a project for Fake News Detection especially for Indians news which are in different languages and different formats.

  1. Fake news as image with no or very less text
  2. Fake news on a blog site
  3. Fake news as Tweets
  4. Fake news in Hindi
  5. Fake news in the watsapp group and shared across.

Need your help on the approach. One approach I can think of is using OCR we can read the content of the post, then search those content in the google. If the news is not present in any of the famous print media then we can tag it as fake. However there can be many challenges in this. What if the print media itself gives any fake news shared by someone.

How to handle the scenario where there is no text in the image but the information shown as image is fake.

How to handle posts written in Hindi. ?

And even if we detect fake news, is there any way to make the person accountable for sharing it. ? I know it is little difficult problem to solve. But is there any work currently done by any company on this. ? Any starting point for me to get into this domain ?

$\endgroup$
  • 1
    $\begingroup$ Particularly challenging, but a good question. $\endgroup$ – prakashchhetri Jan 14 at 4:50
7
$\begingroup$

This is a very ambitious project. First it's important to realize that ML cannot really solve this kind of problem in general, it can only help detect the posts which are likely fake news (see for example this other post about measuring credibility, i.e. the same question seen the other way around).

Assuming you work on the text of the message (I'm not competent about images), the first step would be to collect a corpus of data from all your sources and manually label all of it as fake or not (you could also decide to use a score of "fakeness" for instance). From there you could train a model: at first I would suggest something simple like Naive Bayes (it's the traditional model used for spam detection).

In any case it's important to keep in mind that what the model would learn is not whether the input text is a fake news or not, it would only learn to recognize the tell tale signs: for instance fake news tend to use scary words whereas real news tend to be worded more neutrally. But of course there can real news using scary words and fake news using neutral language, so it's not going to be perfect. In order to really reach the goal the only way would be to have the potential fake news checked by humans, who would be able to give a more informed judgement.

The last part of the question is more a legal question. As far as I know it's often very hard on social media to discover who is really the author, only the police can investigate this but they do it only in very serious cases.

| improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ wonderful explanation on the limitations and challenges. $\endgroup$ – prakashchhetri Jan 14 at 4:49
  • 1
    $\begingroup$ @Erwan. Thank you for the reply. And I do understand that ML may not be able to help here.. But atleast detecting the news as fake/not fake should be doable. However "for instance fake news tend to use scary words whereas real news tend to be worded more neutrally." this does not seem to be relevant for India. Here we have very different levels of fake news. And the way fake news is shared by people on social media is kind of dangerous to the society. And about the legal question, my intention is to find the origin of the news. $\endgroup$ – Akash Jan 14 at 9:01
  • 1
    $\begingroup$ @Akash I understand your goal very well, fake news are a nuisance and sometimes cause very serious problems. I don't know the specific case of fake news in India but in general it's not an easy problem to solve, as I said in my answer. $\endgroup$ – Erwan Jan 14 at 21:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.