Twitter is scrutinising your vocabulary. While the social network takes its fight against aggressive language seriously, the platform also wants to leave room for sarcasm. With this in mind, Twitter has announced improvements in detecting hurtful language versus language that is more casual or slang but not abusive.
Insults and biting jokes are a common practice among some groups of friends. Used in a humorous manner, these words still qualify as insults, especially for Twitter. When it launched a system of alerts against tweets containing language considered aggressive, the platform tended to warn the users responsible for the tweets. However, the platform understood that detecting sarcasm was more complex.
“In early tests, people were sometimes prompted unnecessarily because the algorithms powering the prompts struggled to capture the nuance in many conversations and often didn't differentiate between potentially offensive language, sarcasm, and friendly banter,” Twitter outlined on its blog.
Distinguishing between friendly banter and trolling
Twitter claims that its system now better recognises the difference between aggressive language and friendly colloquialisms, thanks to “improvement to our technology to more accurately detect strong language, including profanity”.
In addition to analysing vocabulary, the platform will now focus more attention on the relationships between users, to better distinguish between true friendships and interactions with “trolls”, those users who endlessly spew hateful comments.
Twitter will be able to determine if a comment is really an insult by analysing the frequency of conversations between two users, whether they follow each other and whether they respond to each other by tweet on a regular basis to determine the nature of their relationship.
“For example, if two accounts follow and reply to each other often, there's a higher likelihood that they have a better understanding of preferred tone of communication,” noted the American social network.
This past February, Twitter deployed an alert system allowing its users to modify a message deemed offensive by the social network. According to the latest feedback, the platform observed that 34% of test users chose to modify their tweet before publishing it or decided not to publish it online.
Eleven percent of them received fewer negative comments after being alerted once. Which suggests that the social network’s policies can have a positive influence and help counter aggressive attitudes expressed on its platform, often criticised for its lack of moderation. – AFP Relaxnews