This folder contains scripts for seeing how accurate our tool is when "trained" on a number of codebases. The script uses 2 sets of vocabularies - one that is built from a number of codebases, one that is built from additional codebases.
- Two versions of
context.txt,design,txtandneither.txt- one that was built from a small subset of codebases and one that serves as the "answer" of the correct term categorisation of all the codebases. - Ensure that the codebases are in the
../../data/your-domain-namefolder.
- Example usage can be found in
__main__.py. Replace the parameters with the name of your domain and the locations of the two vocabulary versions. - From the repo root, run the script with
python -m src.machineLearningExperiment. - The accuracy percentages will be written to
tool-results.csv.