ISAAC Short Answer Scoring Backend

This component started out as a fork of sklearnflask, intended to be used as ML backend for UIMA-based Short Answer Assessment in the ISAAC project (https://www.uni-tuebingen.de/isaac). It evolved into a fully-fledged backend for Short Answer Scoring that can be used as a service with arbitrary frontends over HTTP.

HTTP endpoints

The following endpoints currently exist (more detailed documentation at /docs on the running service):

/fetchStoredModels (GET) - fetch IDs of currently stored SAS models

/trainFromAnswers (POST) - train a new model based on submitted answer data

/predictFromAnswers (POST) - predict score of submitted answers

/wipe_models (GET) - remove all models from service

Docker setup

The service can be built and deployed as a container using the supplied Dockerfile. Run the following from the repo root folder to build the image:

docker build -t isaac/sas-backend ./

To run the image in a container:

docker run --name isaac-sas -p 80:80 ramonziai/isaac-ml-service

Manual/development setup

Installing dependencies

pip install -r requirements.txt

Running the development server

A possible server for development is uvicorn:

pip install uvicorn

The service can be run with the following command:

uvicorn main:app --port 9999 --reload

The --port number can be set to any free port of the developer's choice.
For development purposes it is convenient to make use of the --reload flag to automatically restart the service after changes in the code.