If you’re somewhat far-famed on numerous social networks, possibilities are you are exposed to hate speech in your replies or in your comments. French startup Bodyguard recently launched its app and service in English so that it can hide toxic content from your eyes. It has been available in French for a few years and the company has attracted 50,000 users so far.
“We have developed a engineering that spies hate addres on the internet with a 90% to 95% accuracy and merely 2% of fictitiou positive, ” founder and CEO Charles Cohen told me.
The company has started with a mobile app that anyone can use. After you download the app and connect the app with your favorite social networks, you choose the level of moderation. There are several categories, such as insults, form reproaching, moral provocation, sexual harassment, racism and homophobia. You can select whether it’s a low-grade priority or a top priority for each category.
After that, you don’t have to open the app again. Bodyguard examines replies and observes from its servers and makes a decision whether something is OK. For example, it can hide comments, soften consumers, block customers, etc. When you open Instagram or Twitter again, it’s like those nasty observes never existed.
The app currently patronage Twitter, YouTube, Instagram and Twitch. Unfortunately, it can’t process content on Snapchat and TikTok due to API limitations.
Behind the scenes, most moderation assistances rely heavily on machine learning or keyword-based moderation. Bodyguard has chosen a different approach. It algorithmically cleans up specific comments and tries to analyze the content of a comment contextually. It can determine whether a comment is offensive to you, to a third-party person, to a group of persons, etc.
More recently, the startup has propelled a B2B product. Other companies can use a Bodyguard-powered API to moderate criticisms in real-time on their social stages or in their own apps. The corporation charges its customers exploiting a traditional software-as-a-service approach.
Read more: feedproxy.google.com