Don’t Want To See Spoilers? This AI Tool Will Help You
Harin - Jul 16, 2019
Hate spoilers? Worry no more as there is now an AI tool which can help you detect spoilers even before you come across them.
- This South Korean YouTuber Is The Result Of Deepfake Technology
- YouTube AI Mistakes Black And White In Chess For Racism
- AI Is Being Trained To Identify Faces In The Dark Using Thermal Images
For some people, nothing can be more annoying than accidentally reading or watch a spoiler of your favorite shows and movies even before they can watch them. Today, the internet is full of cinematic spoilers which ruin the whole experience of moviegoers. People then do everything they can to block off all those spoilers. Some even take time off of social media to stay away from recommended videos, news item or random tweets. Worry no more as there is now an AI tool which can help you detect spoilers even before you come across them.
A neural network was trained by researchers to develop an AI tool capable of automatically flagging spoilers in online reviews of TV shows, movies, and books. SpoilerNet, the name of the AI, can be used to create a browser extension to keep all the potential spoilers away.
The neural network was trained using about 1.3 mil book reviews from reviewers which have the “spoiler alert” tag on Goodreads. Goodreads is a social networking site for book lovers where people keep track of what they read as well as sharing their opinions with other readers.
The initial goal of the study, presented at the Association for Computational Linguistics, was to understand how internet users write spoilers and which kind of linguistic patterns a spoiler has.
Many websites like Reddit and IMDb have a “spoilers ahead” warning sign so if a reviewer’s content isn’t spoiler-free, they can manually flag it. But SpoilerNet can save you from coming across spoilers in general.
As the researchers tested the system, it proved to successfully detect spoilers with an accuracy of 89 to 92%. Moreover, researchers also used 16,000 reviews with single sentences of around 880 TV shows to train the system. As a result, it could spot TV shows’ spoilers with 74 to 80% of accuracy.
The study presenters hope to use Goodreads dataset to train the system in detecting spoilers in other types of content including Twitter’s posts.
Comments
Sort by Newest | Popular