ct83

ct83 / gpt_detector / 0.2.0

README.md

Overview

This Algorithm allows you to display the probability that the text was generated by OpenAI's GPT Model.

Applicable Scenarios and Problems

OpenAI released the GPT Models which make it extremely easy to generate human-like text.

The text generated using the GPT Models mimics human speech and language patterns which makes it almost indecipherable from normal human text. You can check it out here at Talk to Transformer.

But, luckily OpenAI also released a GPT Detector which makes it possible to detect text generated by the GPT Models.

This algorithm is a hosted version of the same, it allows you to leverage the GPT Detector in your applications with minimal code, no client side configuration.

Usage

Input

Input the text which needs to be analyzed:

ParameterDescription
inputText which needs to be analyzed

It will be ideal if the text is a simple string with minimal punctuations and no emojis, as those are tend to throw the model off sometimes.

Output

Describe the output fields for your algorithm. For example:

ParameterDescription
all tokensNumber of input words
fake_probabilityProbability that the text is fake
real_probabilityProbability that the text is real
used_tokensNumber of words used

Examples

Input

\"Thanatos In appearance, Thanatos was a long, sleek ship.\n\nBattleships The bridge of a Battleship looked like a giant space helmet. The most incredible thing about them was that they could both fire and lay dormant at the same time.\"

Output

{
  "all_tokens": 57,
  "fake_probability": 0.07913803309202194,
  "real_probability": 0.9208620190620422,
  "used_tokens": 57
}