Grammatical Error Correction Web App

A simple web application that accepts input text and corrects it using a natural language processing model.

This is being built from the bottom up as an exercise in model training and deployment.

Planning to implement the following:

• An "error correction" tool that directly fixes input sentences.
• An "error checker" tool that highlights errors for the user's consideration.

Currently, a basic version of the corrector is available, using a sequence-to-sequence model fine-tuned from T5 following the paper "A Simple Recipe for Multilingual Grammatical Error Correction." See the Hugging Face model card and training code for more details.

Setup

# Install redis if necessary, e.g.,
brew install redis

# Create an environment for this project, e.g.,
$ python3 -m venv .venv

# Activate environment, e.g.,
$ . .venv/bin/activate

# Install dependencies:
$ pip install -r requirements.txt

Launching

# Initialize redis
$ redis-server

# Initialize celery worker
$ celery -A app.make_celery.celery_app worker --loglevel=info

# Launch the app
$ python run.py