tico_19.py

# coding=utf-8
# Copyright 2022 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.

import csv
import os
import re
from pathlib import Path
from typing import Dict, List, Tuple

import datasets
from translate.storage.tmx import tmxfile

from seacrowd.utils import schemas
from seacrowd.utils.configs import SEACrowdConfig
from seacrowd.utils.constants import Tasks

_CITATION = """\
@inproceedings{anastasopoulos-etal-2020-tico,
    title = "{TICO}-19: the Translation Initiative for {CO}vid-19",
    author = {Anastasopoulos, Antonios  and
      Cattelan, Alessandro  and
      Dou, Zi-Yi  and
      Federico, Marcello  and
      Federmann, Christian  and
      Genzel, Dmitriy  and
      Guzm{\'a}n, Franscisco  and
      Hu, Junjie  and
      Hughes, Macduff  and
      Koehn, Philipp  and
      Lazar, Rosie  and
      Lewis, Will  and
      Neubig, Graham  and
      Niu, Mengmeng  and
      {\"O}ktem, Alp  and
      Paquin, Eric  and
      Tang, Grace  and
      Tur, Sylwia},
    booktitle = "Proceedings of the 1st Workshop on {NLP} for {COVID}-19 (Part 2) at {EMNLP} 2020",
    month = dec,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.nlpcovid19-2.5",
    doi = "10.18653/v1/2020.nlpcovid19-2.5",
}
"""

# We follow ISO639-3 language code (https://iso639-3.sil.org/code_tables/639/data)
_LANGUAGES = ["ind", "ara", "spa", "fra", "hin", "por", "rus", "zho", "eng", "khm", "zlm", "mya", "tgl", "tam"]
_LOCAL = False
_SUPPORTED_LANG_PAIRS = [
    ("ind", "ara"),
    ("ind", "spa"),
    ("ind", "fra"),
    ("ind", "hin"),
    ("ind", "por"),
    ("ind", "rus"),
    ("ind", "zho"),
    ("ind", "eng"),
    ("ara", "ind"),
    ("spa", "ind"),
    ("fra", "ind"),
    ("hin", "ind"),
    ("por", "ind"),
    ("rus", "ind"),
    ("zho", "ind"),
    ("eng", "ind"),
    ("khm", "eng"),
    ("eng", "khm"),
    ("mya", "eng"),
    ("eng", "mya"),
    ("zlm", "eng"),
    ("eng", "zlm"),
    ("tgl", "eng"),
    ("eng", "tgl"),
    ("tam", "eng"),
    ("eng", "tam"),
]

_LANG_CODE_MAP = {"ind": "id", "ara": "ar", "spa": "es-LA", "fra": "fr", "hin": "hi", "por": "pt-BR", "rus": "ru", "zho": "zh", "eng": "en", "khm": "km", "zlm": "ms", "mya": "my", "tgl": "tl", "tam": "ta"}

_DEVTEST_LANG_PAIRS = [_LANG_CODE_MAP[source_lang] + "-" + _LANG_CODE_MAP[target_lang] for (source_lang, target_lang) in _SUPPORTED_LANG_PAIRS if (source_lang == "eng" or target_lang == "eng")]

_DATASETNAME = "tico_19"

_DESCRIPTION = """\
TICO-19 (Translation Initiative for COVID-19) is sampled from a variety of public sources containing
COVID-19 related content, representing different domains (e.g., news, wiki articles, and others). TICO-19
includes 30 documents (3071 sentences, 69.7k words) translated from English into 36 languages: Amharic,
Arabic (Modern Standard), Bengali, Chinese (Simplified), Dari, Dinka, Farsi, French (European), Hausa,
Hindi, Indonesian, Kanuri, Khmer (Central), Kinyarwanda, Kurdish Kurmanji, Kurdish Sorani, Lingala,
Luganda, Malay, Marathi, Myanmar, Nepali, Nigerian Fulfulde, Nuer, Oromo, Pashto, Portuguese (Brazilian),
Russian, Somali, Spanish (Latin American), Swahili, Congolese Swahili, Tagalog, Tamil, Tigrinya, Urdu, Zulu.
"""

_HOMEPAGE = "https://tico-19.github.io"

_LICENSE = "CC0"

_URLS = {"evaluation": "https://tico-19.github.io/data/tico19-testset.zip", "all": "https://tico-19.github.io/data/TM/all.{lang_pairs}.tmx.zip"}

_SUPPORTED_TASKS = [Tasks.MACHINE_TRANSLATION]

_SOURCE_VERSION = "1.0.0"

_SEACROWD_VERSION = "2024.06.20"


def seacrowd_config_constructor(lang_source, lang_target, schema, version):
    """Construct SEACrowdConfig with tico_19_{lang_source}_{lang_target}_{schema} as the name format"""
    if schema != "source" and schema != "seacrowd_t2t":
        raise ValueError(f"Invalid schema: {schema}")

    if lang_source == "" and lang_target == "":
        return SEACrowdConfig(
            name="tico_19_{schema}".format(schema=schema),
            version=datasets.Version(version),
            description="tico_19 {schema} schema for default language pair (eng-ind)".format(schema=schema),
            schema=schema,
            subset_id="tico_19",
        )
    else:
        return SEACrowdConfig(
            name="tico_19_{src}_{tgt}_{schema}".format(src=lang_source, tgt=lang_target, schema=schema),
            version=datasets.Version(version),
            description="tico_19 {schema} schema for {src}-{tgt} language pair".format(src=lang_source, tgt=lang_target, schema=schema),
            schema=schema,
            subset_id="tico_19",
        )


class Tico19(datasets.GeneratorBasedBuilder):
    """TICO-19 is MT dataset sampled from a variety of public sources containing COVID-19 related content"""

    SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
    SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)

    BUILDER_CONFIGS = [seacrowd_config_constructor(src, tgt, schema, version) for src, tgt in [("", "")] + _SUPPORTED_LANG_PAIRS for schema, version in zip(["source", "seacrowd_t2t"], [_SOURCE_VERSION, _SEACROWD_VERSION])]

    DEFAULT_CONFIG_NAME = "tico_19_source"

    def _info(self) -> datasets.DatasetInfo:
        if self.config.schema == "source":
            features = datasets.Features(
                {
                    "sourceLang": datasets.Value("string"),
                    "targetLang": datasets.Value("string"),
                    "sourceString": datasets.Value("string"),
                    "targetString": datasets.Value("string"),
                    "stringID": datasets.Value("string"),
                    "url": datasets.Value("string"),
                    "license": datasets.Value("string"),
                    "translatorId": datasets.Value("string"),
                }
            )
        elif self.config.schema == "seacrowd_t2t":
            features = schemas.text2text_features

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        """Returns SplitGenerators."""

        try:
            lang_pairs_config = re.search("tico_19_(.+?)_(source|seacrowd_t2t)", self.config.name).group(1)
            lang_src, lang_tgt = lang_pairs_config.split("_")
        except AttributeError:
            lang_src, lang_tgt = "eng", "ind"

        lang_pairs = _LANG_CODE_MAP[lang_src] + "-" + _LANG_CODE_MAP[lang_tgt]

        # dev & test split only applicable to eng-[sea language] language pair
        if lang_pairs in set(_DEVTEST_LANG_PAIRS):
            lang_sea = _LANG_CODE_MAP[lang_tgt] if lang_src == "eng" else _LANG_CODE_MAP[lang_src]

            data_dir = dl_manager.download_and_extract(_URLS["evaluation"])
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TEST,
                    gen_kwargs={"filepath": os.path.join(data_dir, "tico19-testset", "test", f"test.en-{lang_sea}.tsv"), "lang_source": lang_src, "lang_target": lang_tgt},
                ),
                datasets.SplitGenerator(
                    name=datasets.Split.VALIDATION,
                    gen_kwargs={"filepath": os.path.join(data_dir, "tico19-testset", "dev", f"dev.en-{lang_sea}.tsv"), "lang_source": lang_src, "lang_target": lang_tgt},
                ),
            ]
        else:
            data_dir = dl_manager.download_and_extract(_URLS["all"].format(lang_pairs=lang_pairs))
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={"filepath": os.path.join(data_dir, f"all.{lang_pairs}.tmx"), "lang_source": lang_src, "lang_target": lang_tgt},
                )
            ]

    def _generate_examples(self, filepath: Path, lang_source: str, lang_target: str) -> Tuple[int, Dict]:
        """Yields examples as (key, example) tuples."""

        if self.config.schema == "source":
            # eng-[sea language] language pair dataset provided in .tsv format
            if f"{_LANG_CODE_MAP[lang_source]}-{_LANG_CODE_MAP[lang_target]}" in set(_DEVTEST_LANG_PAIRS):
                with open(filepath, encoding="utf-8") as f:
                    reader = csv.reader(f, delimiter="\t", quotechar='"')
                    for id_, row in enumerate(reader):
                        if id_ == 0:
                            continue
                        if lang_source == "eng":
                            source_lang = row[0]
                            target_lang = row[1]
                            source_string = row[2]
                            target_string = row[3]
                        else:
                            source_lang = row[1]
                            target_lang = row[0]
                            source_string = row[3]
                            target_string = row[2]
                        yield id_, {
                            "sourceLang": source_lang,
                            "targetLang": target_lang,
                            "sourceString": source_string,
                            "targetString": target_string,
                            "stringID": row[4],
                            "url": row[5],
                            "license": row[6],
                            "translatorId": row[7],
                        }

            # all language pairs except eng-ind dataset provided in .tmx format
            else:
                with open(filepath, "rb") as f:
                    tmx_file = tmxfile(f)

                for id_, node in enumerate(tmx_file.unit_iter()):
                    try:
                        url = [text for text in node.xmlelement.itertext("prop")][0]
                    except Exception:
                        url = ""
                    yield id_, {
                        "sourceLang": _LANG_CODE_MAP[lang_source],
                        "targetLang": _LANG_CODE_MAP[lang_target],
                        "sourceString": node.source,
                        "targetString": node.target,
                        "stringID": node.getid(),
                        "url": url,
                        "license": "",
                        "translatorId": "",
                    }

        elif self.config.schema == "seacrowd_t2t":
            if f"{_LANG_CODE_MAP[lang_source]}-{_LANG_CODE_MAP[lang_target]}" in set(_DEVTEST_LANG_PAIRS):
                with open(filepath, encoding="utf-8") as f:
                    reader = csv.reader(f, delimiter="\t", quotechar='"')
                    for id_, row in enumerate(reader):
                        if id_ == 0:
                            continue
                        if lang_source == "eng":
                            source_string = row[2]
                            target_string = row[3]
                        else:
                            source_string = row[3]
                            target_string = row[2]
                        yield id_, {"id": row[4], "text_1": source_string, "text_2": target_string, "text_1_name": lang_source, "text_2_name": lang_target}
            else:
                with open(filepath, "rb") as f:
                    tmx_file = tmxfile(f)

                for id_, node in enumerate(tmx_file.unit_iter()):
                    yield id_, {"id": node.getid(), "text_1": node.source, "text_2": node.target, "text_1_name": lang_source, "text_2_name": lang_target}