You are currently viewing GLTR

GLTR

What do you think about this tool?

GLTR is a forensic analysis tool developed by the MIT-IBM Watson AI lab and HarvardNLP that can detect automatically generated text. It recognizes when a text has been made artificially by examining how likely it is that the text was generated by a language model. GLTR graphically evaluates the output of OpenAI’s GPT-2 117M language model, allowing it to rank each word according on how likely it is to have been produced by the model. The most likely words are highlighted in green, followed by yellow and red, and the remaining words are highlighted in purple. GLTR provides a straightforward visual indicator of how likely each word was under the model, making computer-generated text easier to spot.


Visit website

Spread the love

Leave a Reply