Experts doubt Artificial Intelligence efficacy in correcting profanity on Twitter

Chances are that posts might turn out to be bizarre as the algorithm changes as the word a** to face .

Update: 2018-06-05 20:10 GMT
AI is not yet developed to understand some basic statements in our regional languages, said a government official. (Photo: Pixabay)

Hyderabad: If you are tired of convincing your phone that you did not really mean to type “duck” for “f**k”, brace yourself because now Artificial Intelligence has taken on the responsibility of correcting profanity on Twitter and Reddit. Chances are that posts might turn out to be bizarre as the algorithm changes as the word “a**” to “face”.

However, since the algorithm searches for specific words, it will be difficult to analyse if a person’s post is belittling or sarcastic.

“People on social media come up with creative ways of using cuss words. Some use the combination emoticons like symbols of fingers and words. So I wonder how far this algorithm will work,” said P. Pradeep, social media manager of a popular news website.

“AI is not yet developed to understand some basic statements in our regional languages. The feedback is still limited in Indian languages as the software cannot scan in local languages. This research may be limited to English,” said a government official. 

“In our paper, we present experiments using social media texts in English language only. However, the system can be easily trained for any language as long as you have (large) sets of offensive and non-offensive texts,” said Cicero Nogueira dos santos co-author of the research paper. 

Similar News