Machine learning and Artificial Intelligence can address some of the most burning issues out there. The technology is already being leveraged to make things better in a variety of places, including healthcare, security and many more. Google’s Internet safety incubator Jigsaw, has now set out to deploy the tech to take upon a really pressing issue encountered while we are browsing the Internet.
So imagine this: You are reading something really interesting associated with a trending, controversial topic on the Internet. You really enjoyed reading the article and now want to discuss your own perspective with other folks. You head over to the comments section and start engaging with others. It is all going well and there is a healthy sharing of opinions going on but suddenly, some obnoxious jerk jumps in and starts disagreeing with you. Strongly. Using foul language. And you are put off from what could have been an interesting sharing of knowledge and perspective. Welcome to the Internet.
The problem is more serious than you would imagine. For exmample, popular media outlet Reuters have deleted their comment sections due to this very own problem. Out own comments feed is sometimes encounters irrelevant or obnoxious comments, which is why we keep a strict watch over it. Similarly, BuzzFeed has been dallying with curated comments on its platform. So yeah, toxic and irrelevant comments are a major issue that can ruin a person’s article reading experience.
With all these problems in mind, Google has set out to ensure that your discussions and your reading experience is not sabotaged. The Google’s Internet safety incubator Jigsaw, has come up with a brand new, machine-learning powered technology called Perspective, to help keep the bad stuff away from the fore.
Perspective is basically a machine-learning powered AI system that reviews comments and gives each of them a toxicity rating. Since it is intelligent, it can not only recognize slurs and demeaning words, but also attempts to put things in perspective — that is, understand the context in which they are being spoken. If it is relevant and perhaps in someway related to the article in question, the toxicity rating would be lower. However, if its meant to demean someone and is being used as abuse, the toxicity rating would be much higher.
Perspective’s job is limited to assigning these ratings for now. From their on, it is up to the publishers themselves to decide what they want to do with a particular comment that is toxic. They can chose to hide it behind a warning sign so you have to click to open it and aren’t offended by it as you scroll through, or participate in discussions taking place on the comments section.
Publishers can also chose to hide the comments that have a toxicity rating above a particular threshold outright.
Speaking on the topic, Jigsaw president Jared Cohen said:
Imagine trying to have a conversation with your friends about the news you read this morning, but every time you said something, someone shouted in your face, called you a nasty name or accused you of some awful crime. You’d probably leave the conversation. Unfortunately, this happens all too frequently online as people try to discuss ideas on their favorite news sites but instead get bombarded with toxic comments.
What’s more, commenters can also be shown their own toxicity rating before they post (shown a mirror in a manner of speaking) so that they can take a moment to consider and perhaps make changes to their comments.
Jigsaw trained Perspective using the 11,000 plus daily comments received by the New York Times. It also studied harassment on Wikipedia discussions to ensure that it got everything right. It also built upon similar work of the sort done at Yahoo, where the algorithm was able to detect comments that needed moderation with a 90 percent precision.
For now, it can only moderate in English, and you can expect it to make the occasional mistake. But since it is machine learning based, we can expect it to get better with time and improve user experience on the web.
And of course, Google has a brand new product. With the amount of content that gets pushed to the web everyday, and the even greater number of comments that are received over it, we can expects publishers to exhibit great interest in the product. Meanwhile, it also gives us a look at the sheer reach and capability of Artificial Intelligence in addressing problems that would otherwise require huge human effort, with ease.