Yahoo takes a step further!

About:

Online abuse, cyber bullying, and hate speech are rampant when it comes to our social interactions on the internet. People often forget the line between freedom of expression and offence. Many abuse filters fail to spot them as only blacklisted or common insults are detected. Now, the Yahoo team has built an algorithm that could identify these offenders. Yahoo is releasing the first publicly available selected database of online hate speech for fighting abuse, which means that other sites will be able to use Yahoo’s database of comments in order to craft their own algorithms.

How does it work?

Yahoo’s new abuse-detecting algorithm works 90 percent of the time, which makes it more effective than any other filter. Current abuse filters rely on a combination of blacklisted terms, insults, common expressions or phrases and syntax sign to catch hate speech online, but the Yahoo team applied machine learning to their huge repository of reported comments. The technique used is called ‘Word Embedding’. It processes words as vectors rather than simply positive or negative. Yahoo system recognizes an offensive string of words, even if the individual words are harmless on their own.

Yahoo acquired trained comment moderators and also paid some untrained comment moderators to help model the algorithm. They found that the trained moderators were better at finding out which comments were appropriate.

What other measures can be taken to diminish cyber bullying? Please share your thoughts in the comments section below.

Takeaways:

• Many abuse filters fail to spot an abuse as only blacklisted or common insults are detected. Whereas, the Yahoo team has built an algorithm that could identify offensive comments more accurately.
• Yahoo’s new abuse-detecting algorithm claims to work 90 percent of the time.
• The technique used is called ‘Word Embedding’. It processes words as vectors rather than simply positive or negative.

Subscribe with us to get your dose of interesting news, research & opinions in the startup

segment. Fill the form below:

Name

Email*