import json from pathlib import Path from typing import Dict, List, Tuple import datasets import jsonlines from seacrowd.utils.configs import SEACrowdConfig from seacrowd.utils.constants import Licenses, Tasks _CITATION = """\ @article{, author = {Audah, Hanif Arkan and Yuliawati, Arlisa and Alfina, Ika}, title = {A Comparison Between SymSpell and a Combination of Damerau-Levenshtein Distance With the Trie Data Structure}, journal = {2023 10th International Conference on Advanced Informatics: Concept, Theory and Application (ICAICTA)}, volume = {}, year = {2023}, url = {https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10390399&casa_token=HtJUCIGGlWYAAAAA:q8ll1RWmpHtSAq2Qp5uQAE1NJETx7tUYFZIvTO1IWoaYy4eqFETSsm9p6C7tJwLZBGq5y8zc3A&tag=1}, doi = {}, biburl = {https://github.com/ir-nlp-csui/saltik?tab=readme-ov-file#references}, bibsource = {https://github.com/ir-nlp-csui/saltik?tab=readme-ov-file#references} } """ _DATASETNAME = "saltik" _DESCRIPTION = """\ Saltik is a dataset for benchmarking non-word error correction method accuracy in evaluating Indonesian words. It consists of 58,532 non-word errors generated from 3,000 of the most popular Indonesian words. """ _HOMEPAGE = "https://github.com/ir-nlp-csui/saltik" _LANGUAGES = ["ind"] _LICENSE = Licenses.AGPL_3_0.value _LOCAL = False _URLS = { _DATASETNAME: "https://raw.githubusercontent.com/ir-nlp-csui/saltik/main/saltik.json", } _SUPPORTED_TASKS = [] _SOURCE_VERSION = "1.0.0" _SEACROWD_VERSION = "2024.06.20" class Saltik(datasets.GeneratorBasedBuilder): """It consists of 58,532 non-word errors generated from 3,000 of the most popular Indonesian words.""" SOURCE_VERSION = datasets.Version(_SOURCE_VERSION) SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION) BUILDER_CONFIGS = [ SEACrowdConfig( name=f"{_DATASETNAME}_source", version=SOURCE_VERSION, description=f"{_DATASETNAME} source schema", schema="source", subset_id=f"{_DATASETNAME}", ), ] DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source" def _info(self) -> datasets.DatasetInfo: if self.config.schema == "source": # EX: Arbitrary NER type dataset features = datasets.Features( { "id": datasets.Value("string"), "word": datasets.Value("string"), "errors": [ { "typo": datasets.Value("string"), "error_type": datasets.Value("string"), } ], } ) else: raise NotImplementedError() 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.""" urls = _URLS[_DATASETNAME] file_path = dl_manager.download(urls) data = self._read_jsonl(file_path) all_words = list(data.keys()) processed_data = [] id = 0 for word in all_words: processed_data.append({"id": id, "word": word, "errors": data[word]}) id += 1 self._write_jsonl(file_path + ".jsonl", processed_data) return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, # Whatever you put in gen_kwargs will be passed to _generate_examples gen_kwargs={ "filepath": file_path + ".jsonl", "split": "train", }, ), ] def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]: """Yields examples as (key, example) tuples.""" if self.config.schema == "source": i = 0 with jsonlines.open(filepath) as f: for each_data in f.iter(): ex = { "id": each_data["id"], "word": each_data["word"], "errors": each_data["errors"], } yield i, ex i += 1 def _read_jsonl(self, filepath: Path): with open(filepath) as user_file: parsed_json = json.load(user_file) return parsed_json def _write_jsonl(self, filepath, values): with jsonlines.open(filepath, "w") as writer: for line in values: writer.write(line)