Fannotate is a tool for faster text annotation aided by LLMs. Central to Fannotate is the codebook (annotation guidelines) which help the LLM do an initial guess at the categories and labels.
- Upload and manage annotation codebooks
- Auto-annotation using LLMs
- Manual review and correction interface
- Progress tracking and agreement analysis
- Support for both categorical and free-text annotations
With Fannotate, we aim to strike the right balance between quality and quantity by letting the LLM auto-fill a suggested annotation in addition to a summary of the full text, reducing the need to read through every piece of text to annotate.
-
Clone this repository:
git clone https://github.com/tnresearch/fannotate.git cd fannotate -
Create and activate a conda environment:
conda create -n fannotate python=3.10 conda activate fannotate
-
Install dependencies:
pip install -r requirements.txt
-
Run Fannotate:
python -m fannotate
-
Open your browser and navigate to http://localhost:1337
-
Build the Docker image:
docker build -t fannotate . -
Run the container:
docker run -d -p 1337:1337 -v $(pwd)/uploads:/app/uploads fannotate
- The application will be exposed on port 1337
- When using Docker, the
uploadsdirectory is automatically mounted and will keep data between container restarts - All uploaded files will be stored in the
uploadsdirectory - Fannotate will back up tables to the
uploadsdirectory