probability_words_nli.py

# coding=utf-8
# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# Lint as: python3


import csv
import os
import textwrap
import numpy as np
import datasets
import pandas as pd


_CITATION = """\
@article{sileo2022probing,
  title={Probing neural language models for understanding of words of estimative probability},
  author={Sileo, Damien and Moens, Marie-Francine},
  journal={arXiv preprint arXiv:2211.03358},
  year={2022}
}
"""

_DESCRIPTION = """\
Probing neural language models for understanding of words of estimative probability
"""

URL = 'https://sileod.s3.eu-west-3.amazonaws.com/probability_words/'


class WepProbeConfig(datasets.BuilderConfig):
    """BuilderConfig for WepProbe."""

    def __init__(
        self,
        data_dir,
        label_classes=None,
        process_label=lambda x: x,
        **kwargs,
    ):

        super(WepProbeConfig, self).__init__(version=datasets.Version("1.0.5", ""), **kwargs)
        self.text_features = {k:k for k in ['context', 'hypothesis', 'valid_hypothesis', 'invalid_hypothesis','probability_word','distractor','hypothesis_assertion']}
        self.label_column = 'label'
        self.label_classes = ['valid', 'invalid']
        self.data_url = URL
        self.url=URL
        self.data_dir=data_dir
        self.citation = _CITATION
        self.process_label = process_label


class WepProbe(datasets.GeneratorBasedBuilder):
    """Evaluation of word estimative of probability understanding"""

    BUILDER_CONFIGS = [
        WepProbeConfig(
            name="reasoning_1hop",
            data_dir="reasoning_1hop"),
         WepProbeConfig(
            name="reasoning_2hop",
            data_dir="reasoning_2hop"),
        WepProbeConfig(
            name="usnli",
            data_dir="usnli"),
        ]

    def _info(self):
        features = {text_feature: datasets.Value("string") for text_feature in self.config.text_features.keys()}
        if self.config.label_classes:
            features["label"] = datasets.features.ClassLabel(names=self.config.label_classes)
        else:
            features["label"] = datasets.Value("float32")
        features["idx"] = datasets.Value("int32")
        features["probability"] = datasets.Value("float32")

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(features),
            homepage=self.config.url,
            citation=self.config.citation + "\n" + _CITATION,
        )
    def _split_generators(self, dl_manager):
        
        data_dirs=[]
        for split in ['train','validation','test']:
            url=f'{URL}{self.config.data_dir}_{split}.csv'
            print(url)
            data_dirs+=[dl_manager.download(url)]
        print(data_dirs)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_file": data_dirs[0],
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "data_file": data_dirs[1],
                    "split": "dev",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_file": data_dirs[2],
                    "split": "test",
                },
            ),
        ]

    def _generate_examples(self, data_file, split):
        df = pd.read_csv(data_file).drop(['rnd','split','_'],axis=1,errors='ignore')
        df['idx']=df.index
        for idx, example in df.iterrows():
            yield idx, dict(example)