ParchmentProphet-ui is a Streamlit-based application designed to evaluate and compare human-generated and AI-generated text samples. This tool leverages the ParchmentProphet package to compute linguistic and conceptual similarity metrics between human and AI text samples.
A hosted version is available here: https://parchmentprophet-ca.bravesea-a3cf7791.uksouth.azurecontainerapps.io
-
Upload text files containing human and AI-generated text samples for evaluation.
-
Compare the performance of different AI models based on the provided text samples.
-
View summary statistics and error bar plots for various linguistic features.
-
Explore example text generations for the best and worst performing models.
Your uploaded evaluation data must adhere to the following format:
- Plain text format with a .txt extension.
- Each file should contain multiple sections separated by
*****
, representing different models. - Within each model section, pairs of human and AI-generated text should be separated by
[----]
. - Example
Human text 1
[----]
AI text 1
[----]
Human text 2
[----]
AI text 2
*****
Human text 3
[----]
AI text 3
...
- Clone the repository
git clone https://github.com/Sankgreall/ParchmentProphet-ui.git
cd ParchmentProphet-ui
-
Copy
.env.sample
to.env
and fill in the required variables. You will need an OpenAI API key. -
Build the Docker image
docker build -t parchmentprophet-ui .
- Run the Docker container
docker run -p 8501:8501 parchmentprophet-ui
- Access the application on `http://localhost:8501