|
import os |
|
from zipfile import ZipFile |
|
|
|
import datasets |
|
|
|
|
|
_CITATION = """\ |
|
@misc{storks2021tiered, |
|
title={Tiered Reasoning for Intuitive Physics: Toward Verifiable Commonsense Language Understanding}, |
|
author={Shane Storks and Qiaozi Gao and Yichi Zhang and Joyce Chai}, |
|
year={2021}, |
|
booktitle={Findings of the Association for Computational Linguistics: EMNLP 2021}, |
|
location={Punta Cana, Dominican Republic}, |
|
publisher={Association for Computational Linguistics}, |
|
} |
|
""" |
|
|
|
_DESCRIPTION = """\ |
|
We introduce Tiered Reasoning for Intuitive Physics (TRIP), a novel commonsense reasoning dataset with dense annotations that enable multi-tiered evaluation of machines’ reasoning process. |
|
""" |
|
|
|
_HOMEPAGE = "https://huggingface.co/datasets/sled-umich/TRIP" |
|
|
|
SERVER_URL = "http://162.212.153.129/" |
|
|
|
class TRIP(datasets.GeneratorBasedBuilder): |
|
VERSION = datasets.Version("1.0.1") |
|
|
|
def _info(self): |
|
return datasets.DatasetInfo( |
|
description=_DESCRIPTION, |
|
features=datasets.Features( |
|
{ |
|
"example_id": datasets.Value("string"), |
|
"length": datasets.Value("int32"), |
|
"label": datasets.Value("int32"), |
|
"breakpoint": datasets.Value("int32"), |
|
"confl_sents": [datasets.Value("int32")], |
|
"confl_pairs": [[datasets.Value("int32")]], |
|
"stories":[{ |
|
"story_id": datasets.Value("int32"), |
|
"worker_id": datasets.Value("string"), |
|
"type": datasets.Value("string"), |
|
"idx": datasets.Value("int32"), |
|
"aug": datasets.Value("bool"), |
|
"actor": datasets.Value("string"), |
|
"location": datasets.Value("string"), |
|
"objects": datasets.Value("string"), |
|
"sentences": datasets.features.Sequence(datasets.Value("string")), |
|
"length": datasets.Value("int32"), |
|
"example_id": datasets.Value("string"), |
|
"plausible": datasets.Value("bool"), |
|
"breakpoint": datasets.Value("int32"), |
|
"confl_sents": datasets.features.Sequence(datasets.Value("int32")), |
|
"confl_pairs": [[datasets.Value("int32")]], |
|
"state-h_location": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-conscious": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-wearing": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-h_wet": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-hygiene": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-location": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-exist": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-clean": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-power": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-functional": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-pieces": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-wet": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-open": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-temperature": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-solid": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-contain": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-running": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-moveable": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-mixed": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
"state-edible": [[{"entity": datasets.Value("string"), "num": datasets.Value("int32")}]], |
|
}] |
|
} |
|
), |
|
homepage=_HOMEPAGE, |
|
citation=_CITATION, |
|
) |
|
|
|
def _split_generators(self, dl_manager: datasets.DownloadManager): |
|
"""Returns SplitGenerators.""" |
|
splits = ["ClozeDev", "ClozeTest", "ClozeTrain", "OrderDev", "OrderTest", "OrderTrain"] |
|
data_roots = dl_manager.download_and_extract({k: SERVER_URL + f"trip-{k}.jsonl" for k in splits}) |
|
return [ |
|
datasets.SplitGenerator( |
|
name=split, |
|
gen_kwargs={ |
|
"filepath": data_roots[split], |
|
}, |
|
) |
|
for split in splits |
|
] |
|
|
|
def _generate_examples(self, filepath): |
|
|
|
import json |
|
with open(filepath) as f: |
|
data = [json.loads(line) for line in f] |
|
for i, example in enumerate(data): |
|
yield i, example |