detect toxic and disrespectful comments

Google’s New Tool Against Offensive Comments

Google has announced the creation of Perspective, an artificial intelligence tool so that the media can automatically detect toxic and disrespectful comments.

“Perspective is a technology that uses machine learning to identify toxic comments. And for toxicity we mean those comments that lead someone to abandon a conversation, “Jared Cohen, head of the initiative, said in a conference call.

Google’s new artificial intelligence project aims to be a useful tool to combat “online harassment and the increasing toxicity of online conversations.”

“A number of individuals are hampering civilized and dynamic discussion in the platform and media comment sections, leading to many people leaving the conversation. (…) Many media have ended closing these sections, “added Cohen.

The manager is president of Jigsaw, the Alphabet ideas laboratory – known as Google Ideas – that has been responsible for the development of this artificial intelligence project based on Google’s automatic learning technology, Tensorflow.

Perspective analyzes the comments and detects the degree of toxicity of the conversations, offering media and internet platforms detailed information on a score from zero to one hundred.

Cohen pointed out that it is the media who have to decide how to use that marker: for example, to detect comments that need to be reviewed by hand, to show readers that degree of toxicity or to allow them to filter comments based on the level Are willing to tolerate.

And how does the algorithm know which comments are toxic?

The algorithm compares the analyzed comments with others that were previously categorized as toxic. It also has a system that corrects erroneous evaluations. The more information you manage, the more accurate your judgment will be.

Cohen stressed that Perspective is in “its early days” and “far from perfect” but that it will improve as it has more data to train its accuracy.

The technology has already tested the algorithm with “The New York Times”, “The Guardian”, “The Economist” and the Wikimedia Foundation, responsible for Wikipedia.

Any medium or Internet platform can use Perspective for free – through an API – although for the moment there is only one model for the English language.

In the future, as publishers of other languages join the project, Google’s “aspiration” is to expand the range of languages available .

They also hope that throughout the year Perspective evolves and serves to detect in the comments direct attacks and issues that have nothing to do with the ongoing discussion.

1 thought on “Google’s New Tool Against Offensive Comments”

Leave a Reply

Your email address will not be published. Required fields are marked *