unisent.py

# coding=utf-8


from pathlib import Path
from typing import Dict, List, Tuple

import datasets

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

_CITATION = """\
@inproceedings{asgari2020unisent,
title={UniSent: Universal Adaptable Sentiment Lexica for 1000+ Languages},
author={Asgari, Ehsaneddin and Braune, Fabienne and Ringlstetter, Christoph and Mofrad, Mohammad RK},
booktitle={Proceedings of the International Conference on Language Resources and Evaluation (LREC-2020)},
year={2020},
organization={European Language Resources Association (ELRA)}
}
"""
_DATASETNAME = "unisent"
_DESCRIPTION = """\
UniSent is a universal sentiment lexica for 1000+ languages.
To build UniSent, the authors use a massively parallel Bible
corpus to project sentiment information from English to other
languages for sentiment analysis on Twitter data. 173 of 1404
languages are spoken in Southeast Asia
"""
_URLS = "https://raw.githubusercontent.com/ehsanasgari/UniSent/master/unisent_lexica_v1/{}_unisent_lexicon.txt"
_HOMEPAGE = "https://github.com/ehsanasgari/UniSent"
_LANGUAGES = [
 'aaz',
 'abx',
 'ace',
 'acn',
 'agn',
 'agt',
 'ahk',
 'akb',
 'alj',
 'alp',
 'amk',
 'aoz',
 'atb',
 'atd',
 'att',
 'ban',
 'bbc',
 'bcl',
 'bgr',
 'bgs',
 'bgz',
 'bhp',
 'bkd',
 'bku',
 'blw',
 'blz',
 'bnj',
 'bpr',
 'bps',
 'bru',
 'btd',
 'bth',
 'bto',
 'bts',
 'btx',
 'bug',
 'bvz',
 'bzi',
 'cbk',
 'ceb',
 'cfm',
 'cgc',
 'clu',
 'cmo',
 'cnh',
 'cnw',
 'csy',
 'ctd',
 'czt',
 'dgc',
 'dtp',
 'due',
 'duo',
 'ebk',
 'fil',
 'gbi',
 'gdg',
 'gor',
 'heg',
 'hil',
 'hlt',
 'hnj',
 'hnn',
 'hvn',
 'iba',
 'ifa',
 'ifb',
 'ifk',
 'ifu',
 'ify',
 'ilo',
 'ind',
 'iry',
 'isd',
 'itv',
 'ium',
 'ivb',
 'ivv',
 'jav',
 'jra',
 'kac',
 'khm',
 'kix',
 'kje',
 'kmk',
 'kne',
 'kqe',
 'krj',
 'ksc',
 'ksw',
 'kxm',
 'lao',
 'lbk',
 'lew',
 'lex',
 'lhi',
 'lhu',
 'ljp',
 'lsi',
 'lus',
 'mad',
 'mak',
 'mbb',
 'mbd',
 'mbf',
 'mbi',
 'mbs',
 'mbt',
 'mej',
 'mkn',
 'mmn',
 'mnb',
 'mnx',
 'mog',
 'mqj',
 'mqy',
 'mrw',
 'msb',
 'msk',
 'msm',
 'mta',
 'mtg',
 'mtj',
 'mvp',
 'mwq',
 'mwv',
 'mya',
 'nbe',
 'nfa',
 'nia',
 'nij',
 'nlc',
 'npy',
 'obo',
 'pag',
 'pam',
 'plw',
 'pmf',
 'pne',
 'ppk',
 'prf',
 'prk',
 'pse',
 'ptu',
 'pww',
 'sas',
 'sbl',
 'sda',
 'sgb',
 'smk',
 'sml',
 'sun',
 'sxn',
 'szb',
 'tbl',
 'tby',
 'tcz',
 'tdt',
 'tgl',
 'tha',
 'tih',
 'tlb',
 'twu',
 'urk',
 'vie',
 'war',
 'whk',
 'wrs',
 'xbr',
 'yli',
 'yva',
 'zom',
 'zyp']

_LICENSE = Licenses.CC_BY_NC_ND_4_0.value  # cc-by-nc-nd-4.0
_LOCAL = False

_SUPPORTED_TASKS = [Tasks.SENTIMENT_ANALYSIS]

_SOURCE_VERSION = "1.0.0"
_SEACROWD_VERSION = "2024.06.20"


class UniSentDataset(datasets.GeneratorBasedBuilder):
    LABELS = ["NEGATIVE", "POSITIVE"]

    BUILDER_CONFIGS = [
        SEACrowdConfig(
            name=f"{_DATASETNAME}_{lang}_source",
            version=datasets.Version(_SOURCE_VERSION),
            description=_DESCRIPTION, schema="source",
            subset_id=f"{_DATASETNAME}_{lang}"
        )
        for lang in _LANGUAGES
    ] + [
        SEACrowdConfig(
            name=f"{_DATASETNAME}_{lang}_seacrowd_text",
            version=datasets.Version(_SEACROWD_VERSION),
            description=_DESCRIPTION,
            schema="seacrowd_text",
            subset_id=f"{_DATASETNAME}_{lang}"
        )
        for lang in _LANGUAGES
    ]

    def _info(self) -> datasets.DatasetInfo:

        if self.config.schema == "source":
            features = datasets.Features(
                {
                    "word": datasets.Value("string"),
                    "lexicon": datasets.Value("string"),
                }
            )
        elif self.config.schema == "seacrowd_text":
            features = schemas.text_features(label_names=self.LABELS)
        else:
            raise Exception(f"Unsupported schema: {self.config.schema}")

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

    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
        lang = self.config.subset_id.split("_")[-1]
        url = _URLS.format(lang)
        data_dir = dl_manager.download_and_extract(url)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": data_dir,
                },
            ),
        ]

    def _generate_examples(self, filepath: Path) -> Tuple[int, Dict]:
        with open(filepath, "r", encoding="utf-8") as filein:
            data_instances = [inst.strip("\n").split("\t") for inst in filein.readlines()]

        for di_idx, data_instance in enumerate(data_instances):
            word, lexicon = data_instance
            if self.config.schema == "source":
                yield di_idx, {"word": word, "lexicon": lexicon}
            elif self.config.schema == "seacrowd_text":
                yield di_idx, {"id": di_idx, "text": word, "label": self.LABELS[self._clip_label(int(lexicon))]}
            else:
                raise Exception(f"Unsupported schema: {self.config.schema}")

    @staticmethod
    def _clip_label(label: int) -> int:
        """
        Original labels are -1, +1.
        Clip the label to 0 or 1 to get right index.
        """
        return 0 if int(label) < 0 else 1