gem / gem.py
albertvillanova's picture
Fix URL in gem dataset for totto config (#4396)
696adee
raw
history blame
65.1 kB
# coding=utf-8
# 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.
"""GEM: Generation Evaluation Metrics supporting datasets"""
import csv
import json
import os
import datasets
_CITATION = """\
@article{gem_benchmark,
author = {Sebastian Gehrmann and
Tosin P. Adewumi and
Karmanya Aggarwal and
Pawan Sasanka Ammanamanchi and
Aremu Anuoluwapo and
Antoine Bosselut and
Khyathi Raghavi Chandu and
Miruna{-}Adriana Clinciu and
Dipanjan Das and
Kaustubh D. Dhole and
Wanyu Du and
Esin Durmus and
Ondrej Dusek and
Chris Emezue and
Varun Gangal and
Cristina Garbacea and
Tatsunori Hashimoto and
Yufang Hou and
Yacine Jernite and
Harsh Jhamtani and
Yangfeng Ji and
Shailza Jolly and
Dhruv Kumar and
Faisal Ladhak and
Aman Madaan and
Mounica Maddela and
Khyati Mahajan and
Saad Mahamood and
Bodhisattwa Prasad Majumder and
Pedro Henrique Martins and
Angelina McMillan{-}Major and
Simon Mille and
Emiel van Miltenburg and
Moin Nadeem and
Shashi Narayan and
Vitaly Nikolaev and
Rubungo Andre Niyongabo and
Salomey Osei and
Ankur P. Parikh and
Laura Perez{-}Beltrachini and
Niranjan Ramesh Rao and
Vikas Raunak and
Juan Diego Rodriguez and
Sashank Santhanam and
Joao Sedoc and
Thibault Sellam and
Samira Shaikh and
Anastasia Shimorina and
Marco Antonio Sobrevilla Cabezudo and
Hendrik Strobelt and
Nishant Subramani and
Wei Xu and
Diyi Yang and
Akhila Yerukola and
Jiawei Zhou},
title = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
Metrics},
journal = {CoRR},
volume = {abs/2102.01672},
year = {2021},
url = {https://arxiv.org/abs/2102.01672},
archivePrefix = {arXiv},
eprint = {2102.01672}
}
"""
_DESCRIPTION = """\
GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation,
both through human annotations and automated Metrics.
GEM aims to:
- measure NLG progress across 13 datasets spanning many NLG tasks and languages.
- provide an in-depth analysis of data and models presented via data statements and challenge sets.
- develop standards for evaluation of generated text using both automated and human metrics.
It is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development
by extending existing data or developing datasets for additional languages.
"""
_HOMEPAGE = "https://gem-benchmark.github.io/"
_LICENSE = "CC-BY-SA-4.0"
_TASKS = {
"summarization": {
"mlsum": ["mlsum_de", "mlsum_es"],
"wiki_lingua": [
"wiki_lingua_es_en_v0",
"wiki_lingua_ru_en_v0",
"wiki_lingua_tr_en_v0",
"wiki_lingua_vi_en_v0",
"wiki_lingua_arabic_ar",
"wiki_lingua_chinese_zh",
"wiki_lingua_czech_cs",
"wiki_lingua_dutch_nl",
"wiki_lingua_english_en",
"wiki_lingua_french_fr",
"wiki_lingua_german_de",
"wiki_lingua_hindi_hi",
"wiki_lingua_indonesian_id",
"wiki_lingua_italian_it",
"wiki_lingua_japanese_ja",
"wiki_lingua_korean_ko",
"wiki_lingua_portuguese_pt",
"wiki_lingua_russian_ru",
"wiki_lingua_spanish_es",
"wiki_lingua_thai_th",
"wiki_lingua_turkish_tr",
"wiki_lingua_vietnamese_vi",
],
"xsum": ["xsum"],
},
"struct2text": {
"common_gen": ["common_gen"],
"cs_restaurants": ["cs_restaurants"],
"dart": ["dart"],
"e2e": ["e2e_nlg"],
"totto": ["totto"],
"web_nlg": ["web_nlg_en", "web_nlg_ru"],
},
"simplification": {
"wiki_auto_asset_turk": ["wiki_auto_asset_turk"],
},
"dialog": {
"schema_guided_dialog": ["schema_guided_dialog"],
},
}
_URLs = {
"common_gen": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/common_gen/commongen_data.zip",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/common_gen.zip",
},
"cs_restaurants": {
"train": "https://raw.githubusercontent.com/UFAL-DSG/cs_restaurant_dataset/master/train.json",
"validation": "https://raw.githubusercontent.com/UFAL-DSG/cs_restaurant_dataset/master/devel.json",
"test": "https://raw.githubusercontent.com/UFAL-DSG/cs_restaurant_dataset/master/test.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/cs_restaurants.zip",
},
"dart": {
"train": "https://raw.githubusercontent.com/Yale-LILY/dart/master/data/v1.1.1/dart-v1.1.1-full-train.json",
"validation": "https://raw.githubusercontent.com/Yale-LILY/dart/master/data/v1.1.1/dart-v1.1.1-full-dev.json",
"test": "https://raw.githubusercontent.com/Yale-LILY/dart/master/data/v1.1.1/dart-v1.1.1-full-test.json",
},
"e2e_nlg": {
"train": "https://github.com/tuetschek/e2e-cleaning/raw/master/cleaned-data/train-fixed.no-ol.csv",
"validation": "https://github.com/tuetschek/e2e-cleaning/raw/master/cleaned-data/devel-fixed.no-ol.csv",
"test": "https://github.com/tuetschek/e2e-cleaning/raw/master/cleaned-data/test-fixed.csv",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/e2e_nlg.zip",
},
"mlsum_de": {
"train": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/de_train.zip",
"validation": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/de_val.zip",
"test": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/de_test.zip",
"bad_ids": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_mlsum_bad_ids_fixed.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/mlsum_de.zip",
},
"mlsum_es": {
"train": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/es_train.zip",
"validation": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/es_val.zip",
"test": "https://gitlab.lip6.fr/scialom/mlsum_data/-/raw/master/MLSUM/es_test.zip",
"bad_ids": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_mlsum_bad_ids_fixed.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/mlsum_es.zip",
},
"schema_guided_dialog": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_sgd_context.zip",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/schema_guided_dialog.zip",
},
"totto": {
"data": "https://storage.googleapis.com/totto-public/totto_data.zip",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/totto.zip",
},
"web_nlg_en": {
"train": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_en_train.json",
"validation": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_en_val.json",
"test": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_en_test.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/web_nlg_en.zip",
},
"web_nlg_ru": {
"train": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_ru_train.json",
"validation": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_ru_val.json",
"test": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_web_nlg/webnlg_ru_test.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/web_nlg_ru.zip",
},
"wiki_auto_asset_turk": {
"train": "https://github.com/chaojiang06/wiki-auto/raw/master/wiki-auto/GEM2021/full_with_split/train.tsv",
"validation": "https://github.com/chaojiang06/wiki-auto/raw/master/wiki-auto/GEM2021/full_with_split/valid.tsv",
"test_turk": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_turk_detokenized.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/wiki_auto_asset_turk_train_valid.zip",
},
"wiki_lingua_es_en_v0": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua.zip",
},
"wiki_lingua_ru_en_v0": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua.zip",
},
"wiki_lingua_tr_en_v0": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua.zip",
},
"wiki_lingua_vi_en_v0": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua.zip",
},
"wiki_lingua_arabic_ar": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/arabic.zip",
},
"wiki_lingua_chinese_zh": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/chinese.zip",
},
"wiki_lingua_czech_cs": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/czech.zip",
},
"wiki_lingua_dutch_nl": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/dutch.zip",
},
"wiki_lingua_english_en": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/english.zip",
},
"wiki_lingua_french_fr": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/french.zip",
},
"wiki_lingua_german_de": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/german.zip",
},
"wiki_lingua_hindi_hi": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/hindi.zip",
},
"wiki_lingua_indonesian_id": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/indonesian.zip",
},
"wiki_lingua_italian_it": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/italian.zip",
},
"wiki_lingua_japanese_ja": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/japanese.zip",
},
"wiki_lingua_korean_ko": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/korean.zip",
},
"wiki_lingua_portuguese_pt": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/portuguese.zip",
},
"wiki_lingua_russian_ru": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/russian.zip",
},
"wiki_lingua_spanish_es": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/spanish.zip",
},
"wiki_lingua_thai_th": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/thai.zip",
},
"wiki_lingua_turkish_tr": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/turkish.zip",
},
"wiki_lingua_vietnamese_vi": {
"data": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_wikilingua_full/vietnamese.zip",
},
"xsum": {
"data": "http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz",
"splits": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_xsum_confidence_0.8.json",
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/xsum.zip",
},
}
# Add Asset files
_URLs["wiki_auto_asset_turk"][
"test_asset_orig"
] = "https://github.com/facebookresearch/asset/raw/main/dataset/asset.test.orig"
for i in range(10):
_URLs["wiki_auto_asset_turk"][
f"test_asset_{i}"
] = f"https://github.com/facebookresearch/asset/raw/main/dataset/asset.test.simp.{i}"
_SGD_ACTS = [
"AFFIRM",
"AFFIRM_INTENT",
"CONFIRM",
"GOODBYE",
"INFORM",
"INFORM_COUNT",
"INFORM_INTENT",
"NEGATE",
"NEGATE_INTENT",
"NOTIFY_FAILURE",
"NOTIFY_SUCCESS",
"OFFER",
"OFFER_INTENT",
"REQUEST",
"REQUEST_ALTS",
"REQ_MORE",
"SELECT",
"THANK_YOU",
]
_XSUM_REMOVE_LINES = set(
[
"Share this with\n",
"Email\n",
"Facebook\n",
"Messenger\n",
"Twitter\n",
"Pinterest\n",
"WhatsApp\n",
"Linkedin\n",
"LinkedIn\n",
"Copy this link\n",
"These are external links and will open in a new window\n",
]
)
class Gem(datasets.GeneratorBasedBuilder):
"""GEM: datasets supporting the Generation Evaluation Metrics 2021 shared task."""
BUILDER_CONFIGS = [
datasets.BuilderConfig(
name=conf,
version=datasets.Version("1.1.0"),
description=f"GEM benchmark: {task} task, {conf} subset",
)
for task, dset_confs in _TASKS.items()
for conf_list in dset_confs.values()
for conf in conf_list
]
DEFAULT_CONFIG_NAME = "common_gen" # First alphabetical
def _info(self):
if self.config.name == "common_gen":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"concept_set_id": datasets.Value("int32"),
"concepts": [datasets.Value("string")],
"target": datasets.Value("string"), # single target for train
"references": [datasets.Value("string")], # multiple references for validation
}
)
elif self.config.name == "cs_restaurants":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"dialog_act": datasets.Value("string"),
"dialog_act_delexicalized": datasets.Value("string"),
"target_delexicalized": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name == "dart":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"dart_id": datasets.Value("int32"),
"tripleset": [[datasets.Value("string")]], # list of triples
"subtree_was_extended": datasets.Value("bool"),
"target_sources": [datasets.Value("string")],
"target": datasets.Value("string"), # single target for train
"references": [datasets.Value("string")],
}
)
elif self.config.name == "e2e_nlg":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"meaning_representation": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name.startswith("mlsum"):
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"text": datasets.Value("string"),
"topic": datasets.Value("string"),
"url": datasets.Value("string"),
"title": datasets.Value("string"),
"date": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name == "schema_guided_dialog":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"dialog_acts": [
{
"act": datasets.ClassLabel(names=_SGD_ACTS),
"slot": datasets.Value("string"),
"values": [datasets.Value("string")],
}
],
"context": [datasets.Value("string")],
"dialog_id": datasets.Value("string"),
"service": datasets.Value("string"),
"turn_id": datasets.Value("int32"),
"prompt": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name == "totto":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"totto_id": datasets.Value("int32"),
"table_page_title": datasets.Value("string"),
"table_webpage_url": datasets.Value("string"),
"table_section_title": datasets.Value("string"),
"table_section_text": datasets.Value("string"),
"table": [
[
{
"column_span": datasets.Value("int32"),
"is_header": datasets.Value("bool"),
"row_span": datasets.Value("int32"),
"value": datasets.Value("string"),
}
]
],
"highlighted_cells": [[datasets.Value("int32")]],
"example_id": datasets.Value("string"),
"sentence_annotations": [
{
"original_sentence": datasets.Value("string"),
"sentence_after_deletion": datasets.Value("string"),
"sentence_after_ambiguity": datasets.Value("string"),
"final_sentence": datasets.Value("string"),
}
],
"overlap_subset": datasets.Value("string"),
"target": datasets.Value("string"), # single target for train
"references": [datasets.Value("string")],
},
)
elif self.config.name.startswith("web_nlg"):
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"input": [datasets.Value("string")],
"target": datasets.Value("string"), # single target for train
"references": [datasets.Value("string")],
"category": datasets.Value("string"),
"webnlg_id": datasets.Value("string"),
}
)
elif self.config.name == "wiki_auto_asset_turk":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"source": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name.startswith("wiki_lingua"):
if "v0" in self.config.name:
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"source": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
else:
ln = self.config.name.split("_")[-1]
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"source_aligned": datasets.Translation(languages=[ln, "en"]),
"target_aligned": datasets.Translation(languages=[ln, "en"]),
"source": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
elif self.config.name == "xsum":
features = datasets.Features(
{
"gem_id": datasets.Value("string"),
"gem_parent_id": datasets.Value("string"),
"xsum_id": datasets.Value("string"),
"document": datasets.Value("string"),
"target": datasets.Value("string"),
"references": [datasets.Value("string")],
}
)
return datasets.DatasetInfo(
description=_DESCRIPTION,
features=features,
supervised_keys=None,
homepage=_HOMEPAGE,
license=_LICENSE,
citation=_CITATION,
)
def _split_generators(self, dl_manager):
"""Returns SplitGenerators."""
dl_dir = dl_manager.download_and_extract(_URLs[self.config.name])
if self.config.name == "common_gen":
challenge_sets = [
("challenge_train_sample", "train_common_gen_RandomSample500.json"),
("challenge_validation_sample", "validation_common_gen_RandomSample500.json"),
("challenge_test_scramble", "test_common_gen_ScrambleInputStructure500.json"),
]
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "commongen.train.jsonl"),
"split": "train",
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "commongen.dev.jsonl"),
"split": "validation",
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "commongen.test_noref.jsonl"),
"split": "test",
},
),
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name == "cs_restaurants":
challenge_sets = [
("challenge_train_sample", "train_cs_restaurants_RandomSample500.json"),
("challenge_validation_sample", "validation_cs_restaurants_RandomSample500.json"),
("challenge_test_scramble", "test_cs_restaurants_ScrambleInputStructure500.json"),
]
return [
datasets.SplitGenerator(name=spl, gen_kwargs={"filepath": dl_dir[spl], "split": spl})
for spl in ["train", "validation", "test"]
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name == "dart":
return [
datasets.SplitGenerator(name=spl, gen_kwargs={"filepath": dl_dir[spl], "split": spl})
for spl in ["train", "validation", "test"]
]
elif self.config.name == "e2e_nlg":
challenge_sets = [
("challenge_train_sample", "train_e2e_nlg_RandomSample500.json"),
("challenge_validation_sample", "validation_e2e_nlg_RandomSample500.json"),
("challenge_test_scramble", "test_e2e_nlg_ScrambleInputStructure500.json"),
]
return [
datasets.SplitGenerator(name=spl, gen_kwargs={"filepath": dl_dir[spl], "split": spl})
for spl in ["train", "validation", "test"]
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name.startswith("mlsum"):
lang = self.config.name.split("_")[1]
challenge_sets = [
("challenge_train_sample", f"train_mlsum_{lang}_RandomSample500.json"),
("challenge_validation_sample", f"validation_mlsum_{lang}_RandomSample500.json"),
("challenge_test_covid", f"{lang}_test_covid19_cleaned.jsonl"),
]
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": os.path.join(dl_dir["train"], lang + "_train.jsonl"),
"split": "train",
"lang": lang,
"filepaths": dl_dir["bad_ids"],
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": os.path.join(dl_dir["validation"], lang + "_val.jsonl"),
"split": "validation",
"lang": lang,
"filepaths": dl_dir["bad_ids"],
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": os.path.join(dl_dir["test"], lang + "_test.jsonl"),
"split": "test",
"lang": lang,
"filepaths": dl_dir["bad_ids"],
},
),
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name == "schema_guided_dialog":
challenge_sets = [
("challenge_train_sample", "train_schema_guided_dialog_RandomSample500_reformatted.json"),
("challenge_validation_sample", "validation_schema_guided_dialog_RandomSample500_reformatted.json"),
("challenge_test_backtranslation", "test_schema_guided_dialog_BackTranslation500_reformatted.json"),
(
"challenge_test_bfp02",
"test_schema_guided_dialog_ButterFingersPerturbation_p=0.02_500_reformatted.json",
),
(
"challenge_test_bfp05",
"test_schema_guided_dialog_ButterFingersPerturbation_p=0.05_500_reformatted.json",
),
("challenge_test_nopunc", "test_schema_guided_dialog_WithoutPunctuation500_reformatted.json"),
("challenge_test_scramble", "test_schema_guided_dialog_ScrambleInputStructure500_reformatted.json"),
]
return [
datasets.SplitGenerator(
name=spl, gen_kwargs={"filepath": os.path.join(dl_dir["data"], "gem_sgd.json"), "split": spl}
)
for spl in ["train", "validation", "test"]
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name == "totto":
challenge_sets = [
("challenge_train_sample", "train_totto_RandomSample500.json"),
("challenge_validation_sample", "validation_totto_RandomSample500.json"),
("challenge_test_scramble", "test_totto_ScrambleInputStructure500.json"),
]
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "totto_data/totto_train_data.jsonl"),
"split": "train",
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "totto_data/totto_dev_data.jsonl"),
"split": "validation",
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": os.path.join(dl_dir["data"], "totto_data/unlabeled_totto_test_data.jsonl"),
"split": "test",
},
),
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name.startswith("web_nlg"):
ln = self.config.name.split("_")[-1]
challenge_sets = [
("challenge_train_sample", f"train_web_nlg_{ln}_RandomSample500.json"),
("challenge_validation_sample", f"validation_web_nlg_{ln}_RandomSample500.json"),
("challenge_test_scramble", f"test_web_nlg_{ln}_ScrambleInputStructure500.json"),
]
if ln == "en":
challenge_sets += [("challenge_test_numbers", f"test_web_nlg_{ln}_replace_numbers_500.json")]
return [
datasets.SplitGenerator(name=spl, gen_kwargs={"filepath": dl_dir[spl], "split": spl})
for spl in ["train", "validation", "test"]
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], self.config.name, filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name == "wiki_auto_asset_turk":
challenge_sets = [
("challenge_train_sample", "train_wiki_auto_asset_turk_RandomSample500.json"),
("challenge_validation_sample", "validation_wiki_auto_asset_turk_RandomSample500.json"),
("challenge_test_asset_backtranslation", "test_asset_wiki_auto_asset_turk_BackTranslation.json"),
(
"challenge_test_asset_bfp02",
"test_asset_wiki_auto_asset_turk_ButterFingersPerturbation_p=0.02.json",
),
(
"challenge_test_asset_bfp05",
"test_asset_wiki_auto_asset_turk_ButterFingersPerturbation_p=0.05.json",
),
("challenge_test_asset_nopunc", "test_asset_wiki_auto_asset_turk_WithoutPunctuation.json"),
("challenge_test_turk_backtranslation", "detok_test_turk_wiki_auto_asset_turk_BackTranslation.json"),
(
"challenge_test_turk_bfp02",
"detok_test_turk_wiki_auto_asset_turk_ButterFingersPerturbation_p=0.02.json",
),
(
"challenge_test_turk_bfp05",
"detok_test_turk_wiki_auto_asset_turk_ButterFingersPerturbation_p=0.05.json",
),
("challenge_test_turk_nopunc", "detok_test_turk_wiki_auto_asset_turk_WithoutPunctuation.json"),
]
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": dl_dir["train"],
"split": "train",
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": dl_dir["validation"],
"split": "validation",
},
),
datasets.SplitGenerator(
name="test_asset",
gen_kwargs={
"filepath": "",
"split": "test_asset",
"filepaths": [dl_dir["test_asset_orig"]] + [dl_dir[f"test_asset_{i}"] for i in range(10)],
},
),
datasets.SplitGenerator(
name="test_turk",
gen_kwargs={
"filepath": dl_dir["test_turk"],
"split": "test_turk",
},
),
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], "wiki_auto_asset_turk", filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
elif self.config.name.startswith("wiki_lingua"):
if "v0" in self.config.name:
lang = self.config.name.split("_")[-3]
base_dir = os.path.join(dl_dir["data"], "GEM_data_crosslingual", f"{lang}_en")
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": base_dir,
"split": "train",
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": base_dir,
"split": "val",
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": base_dir,
"split": "test",
},
),
]
else:
lang_name = self.config.name.split("_")[-2]
lang = self.config.name.split("_")[-1]
base_dir = os.path.join(dl_dir["data"], lang_name)
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": base_dir,
"split": "train",
"lang": lang,
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": base_dir,
"split": "val",
"lang": lang,
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": base_dir,
"split": "test",
"lang": lang,
},
),
]
elif self.config.name == "xsum":
challenge_sets = [
("challenge_train_sample", "train_xsum_RandomSample500.json"),
("challenge_validation_sample", "validation_xsum_RandomSample500.json"),
("challenge_test_backtranslation", "test_xsum_BackTranslation500.json"),
("challenge_test_bfp_02", "test_xsum_ButterFingersPerturbation_p=0.02_500.json"),
("challenge_test_bfp_05", "test_xsum_ButterFingersPerturbation_p=0.05_500.json"),
("challenge_test_nopunc", "test_xsum_WithoutPunctuation500.json"),
("challenge_test_covid", "en_test_covid19.jsonl"),
]
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"filepath": dl_dir["splits"],
"split": "train",
"filepaths": os.path.join(dl_dir["data"], "bbc-summary-data"),
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"filepath": dl_dir["splits"],
"split": "validation",
"filepaths": os.path.join(dl_dir["data"], "bbc-summary-data"),
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"filepath": dl_dir["splits"],
"split": "test",
"filepaths": os.path.join(dl_dir["data"], "bbc-summary-data"),
},
),
] + [
datasets.SplitGenerator(
name=challenge_split,
gen_kwargs={
"filepath": os.path.join(dl_dir["challenge_set"], "xsum", filename),
"split": challenge_split,
},
)
for challenge_split, filename in challenge_sets
]
def _generate_examples(self, filepath, split, filepaths=None, lang=None):
"""Yields examples."""
if self.config.name == "common_gen":
if split.startswith("challenge"):
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, encoding="utf-8") as f:
id_ = -1
i = -1
for row in f:
row = row.replace(", }", "}") # Fix possible JSON format error
data = json.loads(row)
concepts = [word for word in data["concept_set"].split("#")]
if split == "train":
i += 1
for scene in data["scene"]:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"concept_set_id": i,
"concepts": concepts,
"target": scene,
"references": [],
}
else:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"concept_set_id": id_,
"concepts": concepts,
"target": "" if split == "test" else data["scene"][0],
"references": [] if split == "test" else data["scene"],
}
elif self.config.name == "cs_restaurants":
if split.startswith("challenge"):
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, encoding="utf8") as f:
data = json.load(f)
for id_, instance in enumerate(data):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"dialog_act": instance["da"],
"dialog_act_delexicalized": instance["delex_da"],
"target": instance["text"],
"target_delexicalized": instance["delex_text"],
"references": [] if split == "train" else [instance["text"]],
}
elif self.config.name == "dart":
with open(filepath, encoding="utf-8") as f:
data = json.loads(f.read())
id_ = -1
i = -1
for example in data:
if split == "train":
i += 1
for annotation in example["annotations"]:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"dart_id": i,
"tripleset": example["tripleset"],
"subtree_was_extended": example.get("subtree_was_extended", None), # some are missing
"target_sources": [annotation["source"] for annotation in example["annotations"]],
"target": annotation["text"],
"references": [],
}
else:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"dart_id": id_,
"tripleset": example["tripleset"],
"subtree_was_extended": example.get("subtree_was_extended", None), # some are missing
"target_sources": [annotation["source"] for annotation in example["annotations"]],
"target": example["annotations"][0]["text"] if len(example["annotations"]) > 0 else "",
"references": [annotation["text"] for annotation in example["annotations"]],
}
elif self.config.name == "e2e_nlg":
if split.startswith("challenge"):
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, encoding="utf-8") as f:
reader = csv.DictReader(f)
for id_, example in enumerate(reader):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"meaning_representation": example["mr"],
"target": example["ref"],
"references": [] if split == "train" else [example["ref"]],
}
elif self.config.name.startswith("mlsum"):
if split in ["train", "validation", "test", "challenge_test_covid"]:
if split == "challenge_test_covid":
bad_ids = {}
else:
bad_ids_dct = json.load(open(filepaths, encoding="utf-8"))
bad_ids = dict((bad_url, True) for _, bad_url in bad_ids_dct[f"{lang}-{split}"])
with open(filepath, encoding="utf-8") as f:
id_ = -1
for line in f:
data = json.loads(line)
if data["url"] in bad_ids:
continue
else:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"text": data["text"],
"target": data["summary"],
"references": [] if split == "train" else [data["summary"]],
"topic": data["topic"],
"url": data["url"],
"title": data["title"],
"date": data["date"],
}
else:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
elif self.config.name == "schema_guided_dialog":
if "challenge" in split:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
examples = json.load(open(filepath, encoding="utf-8"))[split]
for id_, example in enumerate(examples):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"dialog_acts": [
{
"act": act_id,
"slot": slot,
"values": values,
}
for act_id, slot, values in example["da"]
],
"context": example["context"],
"dialog_id": example["dialog_id"],
"service": example["service"],
"turn_id": example["turn_ix"],
"prompt": example["prompt"],
"target": example["target"],
"references": [] if split == "train" else [example["target"]],
}
elif self.config.name == "totto":
if "challenge" in split:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, "r", encoding="utf-8") as json_file:
json_list = list(json_file)
id_ = -1
i = -1
for json_str in json_list:
result = json.loads(json_str)
if split == "train":
i += 1
for sentence in result["sentence_annotations"]:
id_ += 1
response = {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"totto_id": i,
"table_page_title": result["table_page_title"],
"table_webpage_url": result["table_webpage_url"],
"table_section_title": result["table_section_title"],
"table_section_text": result["table_section_text"],
"table": result["table"],
"highlighted_cells": result["highlighted_cells"],
"example_id": str(result["example_id"]),
"overlap_subset": "none",
"sentence_annotations": [sentence],
"references": [],
"target": sentence["final_sentence"],
}
yield id_, response
else:
id_ += 1
response = {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"totto_id": id_,
"table_page_title": result["table_page_title"],
"table_webpage_url": result["table_webpage_url"],
"table_section_title": result["table_section_title"],
"table_section_text": result["table_section_text"],
"table": result["table"],
"highlighted_cells": result["highlighted_cells"],
"example_id": str(result["example_id"]),
"overlap_subset": str(result["overlap_subset"]),
"sentence_annotations": [] if split == "test" else result["sentence_annotations"],
}
response["references"] = [
sentence["final_sentence"] for sentence in response["sentence_annotations"]
]
response["target"] = response["references"][0] if len(response["references"]) > 0 else ""
yield id_, response
elif self.config.name.startswith("web_nlg"):
if "challenge" in split:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
if len(exple) == 0:
continue
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, encoding="utf-8") as f:
examples = json.load(f)
id_ = -1
for example in examples["values"]:
if split == "train":
for target in example["target"]:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"input": example["input"],
"target": target,
"references": [] if split == "train" else example["target"],
"category": example["category"],
"webnlg_id": example["webnlg-id"],
}
else:
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"input": example["input"],
"target": example["target"][0] if len(example["target"]) > 0 else "",
"references": example["target"],
"category": example["category"],
"webnlg_id": example["webnlg-id"],
}
elif self.config.name == "wiki_auto_asset_turk":
if split in ["train", "validation"]:
keys = [
"source",
"target",
]
with open(filepath, encoding="utf-8") as f:
for id_, line in enumerate(f):
values = line.strip().split("\t")
assert len(values) == 2, f"Not enough fields in ---- {line} --- {values}"
example = dict([(k, val) for k, val in zip(keys, values)])
example["gem_id"] = f"{self.config.name}-{split}-{id_}"
example["gem_parent_id"] = example["gem_id"]
example["references"] = [] if split == "train" else [example["target"]]
yield id_, example
elif split == "test_turk":
examples = json.load(open(filepath, encoding="utf-8"))
for id_, example in enumerate(examples):
example["gem_parent_id"] = example["gem_id"]
for k in ["source_id", "target_id"]:
if k in example:
del example[k]
yield id_, example
elif split == "test_asset":
files = [open(f_name, encoding="utf-8") for f_name in filepaths]
for id_, lines in enumerate(zip(*files)):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"target": lines[1].strip(),
"source": lines[0].strip(),
"references": [line.strip() for line in lines[1:]],
}
else:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
for k in ["source_id", "target_id"]:
if k in exple:
del exple[k]
yield id_, exple
elif self.config.name.startswith("wiki_lingua"):
if "v0" in self.config.name:
with open(os.path.join(filepath, f"{split}.src"), encoding="utf-8") as f_in:
with open(os.path.join(filepath, f"{split}.tgt"), encoding="utf-8") as f_out:
for id_, (src, tgt) in enumerate(zip(f_in, f_out)):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"source": src.strip(),
"target": tgt.strip(),
"references": [] if split == "train" else [tgt.strip()],
}
else:
with open(os.path.join(filepath, f"{split}.src.{lang}"), encoding="utf-8") as f_in_ln:
with open(os.path.join(filepath, f"{split}.src.en"), encoding="utf-8") as f_in_en:
with open(os.path.join(filepath, f"{split}.tgt.{lang}"), encoding="utf-8") as f_out_ln:
with open(os.path.join(filepath, f"{split}.tgt.en"), encoding="utf-8") as f_out_en:
for id_, (src_ln, src_en, tgt_ln, tgt_en) in enumerate(
zip(f_in_ln, f_in_en, f_out_ln, f_out_en)
):
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"source_aligned": {lang: src_ln.strip(), "en": src_en.strip()},
"target_aligned": {lang: tgt_ln.strip(), "en": tgt_en.strip()},
"source": src_ln.strip(),
"target": tgt_en.strip(),
"references": [] if split == "train" else [tgt_en.strip()],
}
elif self.config.name == "xsum":
if "challenge" in split:
if "covid" in split:
with open(filepath, encoding="utf-8") as f:
id_ = -1
for line in f:
data = json.loads(line)
id_ += 1
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"xsum_id": data["url"],
"document": data["text"],
"target": data["summary"],
"references": [] if split == "train" else [data["summary"]],
}
else:
exples = json.load(open(filepath, encoding="utf-8"))
if isinstance(exples, dict):
assert len(exples) == 1, "multiple entries found"
exples = list(exples.values())[0]
for id_, exple in enumerate(exples):
exple["gem_parent_id"] = exple["gem_id"]
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
yield id_, exple
else:
with open(filepath, "r", encoding="utf-8") as f:
split_ids = json.load(f)
for id_, i in enumerate(split_ids[split]):
with open(os.path.join(filepaths, i + ".summary"), "r", encoding="utf-8") as f:
text = "".join(
[line for line in f.readlines() if line not in _XSUM_REMOVE_LINES and line.strip()]
)
segs = text.split("[SN]")
yield id_, {
"gem_id": f"{self.config.name}-{split}-{id_}",
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
"xsum_id": i,
"document": segs[8].strip(),
"target": segs[6].strip(),
"references": [] if split == "train" else [segs[6].strip()],
}