venom1270 / essay-grading

Essay grading for Orange

This is an essay grading add-on for Orange data mining.

Installation

• Download and install Orange

• Open Orange

  • Click Options > Add-ons
  • Install Orange3-Text add-on

• Clone this repository

• Open Anaconda prompt (as Admin), navigate to cloned repository, activate Orange environment and install essay grading:

    cd essay-grading
    activate [path_to_Orange_environment] (e.g. "C:\Program Files\Orange")
    pip install -e .
    conda install pytorch torchvision cpuonly -c pytorch
    pip install flair==0.4.5
  

  • If any problems arise during Flair installation, try running the following command before installing Flair

      pip install transformers==2.4.1
    

• In addition to the above, we need to take care of neuralcoref manually. Download neuralcoref from here navigate to it and run:

    cd neuralcoref
    pip install -r requirements.txt
    pip install -e .
  

• You will also need to download SpaCy's english language models:

    python -m spacy download en
    python -m spacy download en_core_web_lg 
    python -m spacy download en_vectors_web_lg 
  

• DONE!

  • Your code changes will take effect immediately or after restarting Orange (if open).

Usage

• Run Orange
• All essay grading widgets are bundled in "Essay grading"
• You can find a few examples by clicking Help > Example Workflows

Remarks

This is an implementation of an essay grading system described in PhD thesis by Kaja Zupanc.

AGE and AGE+ systems are fully functional. SAGE is functional but requires a lot of time to complete.