The Stack Archive News Article

New system detects hate speech in Tweets via Deep Learning

Wed 26 Apr 2017

IIIT-Hyderabad’s Informational and Retrieval Extraction Lab (IREL) has devised a system that can detect the presence of hatred in Tweets

The majority of tweets are sent as communication, advertising or offering kind words. Unfortunately, a minority are sent with more nefarious intentions, to spread hate.

Could these kinds of tweet become things of the past? A new method of detecting elements of hatred in tweets has been devised. IIIT-Hyderabad’s Informational and Retrieval Extraction Lab (IREL) has looked into this problem and come up with an automated system that uses Artificial Intelligence chatterbots to detect the presence of hate speech in tweets.

These chatterbots can pick up on instances of racist or sexist speech, abusive language and flag offensive content. Another notable aspect of this new system is to analyse public sentiment, which can then find the cause of the problems via user generated content.

In order to detect the presence of hate speech, an approach known as Supervised Learning is used – a computer algorithm that is fed a number of examples of text used in tweets of a racist or sexist nature.

This algorithm is designed to ‘learn’ when it looks at data. After the algorithm has ended, the programme can then detect the presence of racist or sexist language in text. It uses neural networks (also known as Deep Learning). These draw inspiration from the human brain and attempt to simulate how humans learn from examples.

However, the system does throw up some challenges and raises some questions, including how natural language processing can decipher the various forms of hatred, and how it can identify the targets of hatred.

Nevertheless, the new system has received a warm welcome. Vasudeva Varma, Professor and Dean (R & D) at IIIT-Hyderabad, students Pinkesh Badjatiya, Shashank Gupta and adjunct faculty Manish Gupta presented this at web conference, WWW2017 Perth earlier this month. Their poster on Deep Learning for Hate Speech Detection in Twitter was subsequently voted as the best poster presentation among 166 submissions worldwide.


AI Big Data India news Twitter
Send us a correction about this article Send us a news tip