Instagram starts using AI to reduce bullying

The new system uses DeepText, a text classification motor that has Facebook and Instagram have worked with over past nine months.

Wednesday 12, July 2017 18:07

The tool helps systems to understand how words are used in different contexts with the priority of identifying spam.

The system is the newest tech app designed to reduce spam, as much on social media networks as in emails, and has been in development since last October.

Based on the success of the last system, the Facebook team wanted to try to see if it was possible to identify negative, malicious, racist, insulting or bullying text left in comments or in private messages.

The first trials analysed at least two million comments before the filter was put to work. Each message was verified at least twice by workers to check the functionality of the system.

To date, the filter has only been put into practice in English. However, everyone involved in the project speak at least two languages to native level, and Instagram is hoping to expand their unwanted message filter to other languages. That will mean that the tool will soon be available in other regions.