scue

scue pronounced as skew is a library for analyzing superficial cues in natural language understanding datasets.

Usage

scue supports

• Python 3.7 and up (CPython and PyPy)
• Rust 1.56 and up

import argparse

from scue.models import (
    UnbalancedNgramModelForMultipleChoice,
    SequenceLengthModelForMultipleChoice,
    LexicalOverlapModelForMultipleChoice,
    EnsembleModelForMultipleChoice,
)
from scue.preprocessors import MultipleChoicePreprocessor


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--train_file", type=str, required=True)
    parser.add_argument("--validation_file", type=str, required=False)
    parser.add_argument("--n", type=int, required=True)
    parser.add_argument("--id_field", type=str, required=True)
    parser.add_argument("--context_field", type=str, required=True)
    parser.add_argument("--choices_field", type=str, required=True)
    parser.add_argument("--question_field", type=str, required=False)
    args = parser.parse_args()

    processor = MultipleChoicePreprocessor(
        id_field=args.id_field,
        context_field=args.context_field,
        choices_field=args.choices_field,
        remove_stopwords=False,
        remove_punctuation=False,
    )
    data = processor(args.train_file)

    model = UnbalancedNgramModelForMultipleChoice(n=args.n, training_data=data)
    model.fit()

    # accuracy on training set
    print(model.evaluate(data))

    # print important features
    print(model.important_features)

    if args.validation_file:
        val_data = processor(args.validation_file)
        print(model.evaluate(val_data))

Development

scue requires

• Python 3.7 and up (CPython and PyPy).
• Rust 1.56 and up

First, follow the commands below to create a new directory containing a new Python virtualenv, and install maturin into the virtualenv using Python's package manager, pip:

$ git clone
$ cd scue
$ mkdir string_sum
$ cd string_sum
$ python -m venv .env
$ source .env/bin/activate
$ pip install maturin

Finally, run maturin develop. This will build the rust buckend package and install it into the Python virtualenv previously created and activated. The package is then ready to be used from python:

$ maturin develop

License

scue is licensed under the Apache-2.0 license.