OCW / OCW.py
salavina's picture
Update OCW.py
0e50dc2
raw
history blame
11.1 kB
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# 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.
"""Only Connect Wall (OCW) dataset"""
import json
import os
import datasets
_CITATION = """\
@article{Naeini2023LargeLM,
title = {Large Language Models are Fixated by Red Herrings: Exploring Creative Problem Solving and Einstellung Effect using the Only Connect Wall Dataset},
author = {Saeid Alavi Naeini and Raeid Saqur and Mozhgan Saeidi and John Giorgi and Babak Taati},
year = 2023,
journal = {ArXiv},
volume = {abs/2306.11167},
url = {https://api.semanticscholar.org/CorpusID:259203717}
}
"""
_DESCRIPTION = """\
The Only Connect Wall (OCW) dataset contains 618 "Connecting Walls" from the Round 3: Connecting Wall segment of the Only Connect quiz show, collected from 15 seasons' worth of episodes. Each wall contains the ground-truth groups and connections as well as recorded human performance.
"""
_HOMEPAGE_URL = "https://github.com/TaatiTeam/OCW/"
_LICENSE = "MIT"
_BASE_URL = "https://www.cs.toronto.edu/~taati/OCW/"
_URLS = {
"ocw": _BASE_URL + "OCW.tar.gz",
"ocw_randomized": _BASE_URL + "OCW_randomized.tar.gz",
"ocw_wordnet": _BASE_URL + "OCW_wordnet.tar.gz"
}
class OCW(datasets.GeneratorBasedBuilder):
"""OCW dataset"""
VERSION = datasets.Version("1.0.0")
BUILDER_CONFIGS = [
datasets.BuilderConfig(name="ocw", version=VERSION,
description="main OCW dataset"),
datasets.BuilderConfig(name="ocw_randomized", version=VERSION,
description="Easy OCW dataset with randomized groups in each wall"),
datasets.BuilderConfig(name="ocw_wordnet", version=VERSION,
description="Easy OCW dataset with wordnet synonyms replaced with original clues")
]
DEFAULT_CONFIG_NAME = "ocw"
def _info(self):
features = datasets.Features(
{
# "total_walls_in_season": datasets.Value("int32"),
# "season_start_date": datasets.Value("string"),
# "season_end_date": datasets.Value("string"),
"wall_id": datasets.Value("string"),
"season": datasets.Value("int32"),
"episode": datasets.Value("int32"),
"words": datasets.features.Sequence(feature=datasets.Value("string")),
"gt_connections": datasets.features.Sequence(feature=datasets.Value("string")),
"group_1":
{
"group_id": datasets.Value("string"),
"gt_words":datasets.features.Sequence(feature=datasets.Value("string")),
"gt_connection": datasets.Value("string"),
"human_performance":
{
"grouping": datasets.Value("int32"),
"connection": datasets.Value("int32")
}
},
"group_2":
{
"group_id": datasets.Value("string"),
"gt_words": datasets.features.Sequence(feature=datasets.Value("string")),
"gt_connection": datasets.Value("string"),
"human_performance":
{
"grouping": datasets.Value("int32"),
"connection": datasets.Value("int32")
}
},
"group_3":
{
"group_id": datasets.Value("string"),
"gt_words": datasets.features.Sequence(feature=datasets.Value("string")),
"gt_connection": datasets.Value("string"),
"human_performance":
{
"grouping": datasets.Value("int32"),
"connection": datasets.Value("int32")
}
},
"group_4":
{
"group_id": datasets.Value("string"),
"gt_words": datasets.features.Sequence(feature=datasets.Value("string")),
"gt_connection": datasets.Value("string"),
"human_performance":
{
"grouping": datasets.Value("int32"),
"connection": datasets.Value("int32")
}
},
}
)
return datasets.DatasetInfo(
# This is the description that will appear on the datasets page.
description=_DESCRIPTION,
# This defines the different columns of the dataset and their types
features= features,
# Homepage of the dataset for documentation
homepage=_HOMEPAGE_URL,
# License for the dataset if available
license=_LICENSE,
# Citation for the dataset
citation=_CITATION,
# No default supervised_keys
supervised_keys=None
)
def _split_generators(self, dl_manager):
url = _URLS[self.config.name]
if self.config.name == "ocw_randomized":
url = [url, _URLS[self.DEFAULT_CONFIG_NAME]]
path = dl_manager.download_and_extract(url)
if self.config.name == self.DEFAULT_CONFIG_NAME:
dir = 'dataset'
train_filepath = os.path.join(path, dir, 'train.json')
val_filepath = os.path.join(path, dir, 'validation.json')
test_filepath = os.path.join(path, dir, 'test.json')
elif self.config.name == "ocw_randomized":
# OCW-randomized only contains a test set, we load main OCW train/validation files
dir = 'OCW_randomized'
dir2 = 'dataset'
train_filepath = os.path.join(path[1], dir2, 'train.json')
val_filepath = os.path.join(path[1], dir2, 'validation.json')
test_filepath = os.path.join(path[0], dir, 'easy_test.json')
else:
dir = 'OCW_wordnet'
train_filepath = os.path.join(path, dir, 'easy_train_wordnet.json')
val_filepath = os.path.join(path, dir, 'easy_validation_wordnet.json')
test_filepath = os.path.join(path, dir, 'easy_test_wordnet.json')
return [
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": train_filepath}),
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": val_filepath}),
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": test_filepath}),
]
def _generate_examples(self, filepath):
"""This function returns the examples in the raw (text) form."""
key = 0
with open(filepath, encoding="utf-8") as f:
ocw = json.load(f)
for data in ocw["dataset"]:
wall_id = data.get("wall_id")
season = data.get("season")
# season_to_walls_map = ocw['season_to_walls_map'][str(season)]
# total_walls_in_season = season_to_walls_map["num_walls"]
# season_start_date = season_to_walls_map["start_date"]
# season_end_date = season_to_walls_map["end_date"]
episode = data.get("episode")
words = data.get("words")
gt_connections = data.get("gt_connections")
group_1 = data['groups']['group_1']
group_1_human_performance = group_1['human_performance']
group_2 = data['groups']['group_2']
group_2_human_performance = group_2['human_performance']
group_3 = data['groups']['group_3']
group_3_human_performance = group_3['human_performance']
group_4 = data['groups']['group_4']
group_4_human_performance = group_4['human_performance']
yield key, {
# "total_walls_in_season": total_walls_in_season,
# "season_start_date": season_start_date,
# "season_end_date": season_end_date,
"wall_id": wall_id,
"season": season,
"episode": episode,
"words": words,
"gt_connections": gt_connections,
"group_1": {
"group_id": group_1.get("group_id"),
"gt_words": group_1.get("gt_words"),
"gt_connection": group_1.get("gt_connection"),
"human_performance": {
"grouping": group_1_human_performance.get("grouping"),
"connection": group_1_human_performance.get("connection")
}
},
"group_2": {
"group_id": group_2.get("group_id"),
"gt_words": group_2.get("gt_words"),
"gt_connection": group_2.get("gt_connection"),
"human_performance": {
"grouping": group_2_human_performance.get("grouping"),
"connection": group_2_human_performance.get("connection")
}
},
"group_3": {
"group_id": group_3.get("group_id"),
"gt_words": group_3.get("gt_words"),
"gt_connection": group_3.get("gt_connection"),
"human_performance": {
"grouping": group_3_human_performance.get("grouping"),
"connection": group_3_human_performance.get("connection")
}
},
"group_4": {
"group_id": group_4.get("group_id"),
"gt_words": group_4.get("gt_words"),
"gt_connection": group_4.get("gt_connection"),
"human_performance": {
"grouping": group_4_human_performance.get("grouping"),
"connection": group_4_human_performance.get("connection")
}
},
}
key += 1