# 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. # TODO: Address all TODOs and remove all explanatory comments """Super-EURLEX dataset containing legal documents from multiple languages""" import numpy as np import pandas as pd import datasets # TODO: Add BibTeX citation # Find for instance the citation on arxiv or on the dataset repo/website _CITATION = """ """ # TODO: Add description of the dataset here # You can copy an official description _DESCRIPTION = """ Super-EURLEX dataset containing legal documents from multiple languages. The datasets are build/scrapped from the EURLEX Website [https://eur-lex.europa.eu/homepage.html] With one split per language and sector, because the available features (metadata) differs for each sector. Therefore, each sample contains the content of a full legal document in up to 3 different formats. Those are raw HTML and cleaned HTML (if the HTML format was available on the EURLEX website during the scrapping process) and cleaned text. The cleaned text should be available for each sample and was extracted from HTML or PDF. 'Cleaned' HTML stands here for minor cleaning that was done to preserve to a large extent the necessary HTML information like table structures while removing unnecessary complexity which was introduced to the original documents due to actions like writing each sentence into a new object. Additionally, each sample contains metadata which was scrapped on the fly, this implies the following 2 things. First, not every sector contains the same metadata. Second, most metadata might be irrelevant for most use cases. In our minds the most interesting metadata is the celex-id which is used to identify the legal document at hand, but also contains a lot of information about the document see [https://eur-lex.europa.eu/content/tools/eur-lex-celex-infographic-A3.pdf] as well as eurovoc- concepts, which are labels that define the content of the documents. Eurovoc-Concepts are, for example, only available for the sectors 1, 2, 3, 4, 5, 6, 9, C, and E. The Naming of most metadata is kept like it was on the eurlex website, except for converting it to lower case and replacing whitespaces with '_'. """ # TODO: Add a link to an official homepage for the dataset here _HOMEPAGE = "" # TODO: Add the licence for the dataset here if you can find it _LICENSE = "" AVAILABLE_LANGUAGES="BG CS DA DE EL EN ES ET FI FR GA HR HU IT LT LV MT NL PL PT RO SK SL SV".split(" ") SECTORS=['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'C', 'E'] VERSIONS=[None, 'html']#, 'clean'] # Features to override the standard most Features were scrapped as a sequence # of strings with only the following exceptions FEATURES = { 'celex_id': datasets.Value("string"), 'text_cleaned': datasets.Value("string"), 'text_html_cleaned': datasets.Value("string"), 'text_html_raw': datasets.Value("string"), } FEATURES_IN_SECTOR={ '0':['form'], '1':['form', 'subject_matter', 'current_consolidated_version', 'directory_code', 'eurovoc',], '2':['form', 'directory_code', 'subject_matter', 'eurovoc', 'current_consolidated_version', 'latest_consolidated_version', ], '3':['form', 'directory_code', 'subject_matter', 'eurovoc', 'latest_consolidated_version', 'current_consolidated_version', ], '4':['form', 'eurovoc', 'subject_matter', 'current_consolidated_version', 'directory_code', 'latest_consolidated_version', ], '5':['form', 'directory_code', 'subject_matter', 'eurovoc', 'current_consolidated_version', ], '6':['form', 'case-law_directory_code_before_lisbon', 'subject_matter', 'eurovoc', 'directory_code', ], '7':['form', 'transposed_legal_acts'],#, 'text_html_raw', 'text_html_cleaned', '8':['form', 'case-law_directory_code_before_lisbon', 'subject_matter',], '9':['form', 'eurovoc', 'subject_matter', 'directory_code',], 'C':['form', 'eurovoc'], 'E':['form', 'directory_code', 'subject_matter', 'eurovoc',], } VERSION_FEATURES={ None: ['celex_id', 'text_cleaned'], 'clean': ['celex_id', 'text_cleaned'], 'html': ['celex_id', 'text_html_raw'], } available_features_tmp = [] for version in VERSIONS: v = '' if version is None else f"_{version}" available_features_tmp.append( {sector+v: datasets.Features({feature:(FEATURES[feature] if feature in FEATURES else datasets.Sequence(datasets.Value("string"))) for feature in VERSION_FEATURES[version] + FEATURES_IN_SECTOR[sector]}) for sector in SECTORS} ) # Combine Features Lists into single dictionary with the Features of each version of each sector AVAILABLE_FEATURES={k: v for d in available_features_tmp for k, v in d.items()} SECTOR_DESCRIPTIONS={ '0': "Consolidated acts ", '1': "Treaties", '2': "International agreements", '3': "Legislation", '4': "Complementary legislation", '5': "Preparatory acts and working documents", '6': "Case-law", '7': "National transposition measures", '8': "References to national case-law concerning EU law", '9': "Parliamentary questions", 'C': "Other documents published in the Official Journal C series", 'E': "EFTA documents", } class SuperEurlexConfig(datasets.BuilderConfig): """BuilderConfig for SuperGLUE.""" def __init__(self, sector, language, feature_version, features, citation, url, **kwargs): """BuilderConfig for SuperGLUE. Args: sector: sector of the wanted data language: the language code for the language in which the text shall be written in features: *list[string]*, list of the features that will appear in the feature dict. citation: *string*, citation for the data set. url: *string*, url for information about the data set. **kwargs: keyword arguments forwarded to super. """ name=sector+'.'+language+('' if feature_version==None else f".{feature_version}") super().__init__(name=name, version=datasets.Version("0.1.0"), **kwargs) self.features = features self.language = language self.sector = sector self.feature_version = str(feature_version) self.text_data_url = f"text_data/{language}/{sector}_{'clean' if feature_version is None else feature_version}.parquet" self.meta_data_url = f"meta_data/{sector}.parquet" self.citation = citation self.url = url class FileNotFoundException(Exception): pass # TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case class SuperEurlex(datasets.GeneratorBasedBuilder): """TODO: Short description of my dataset.""" VERSION = datasets.Version("1.1.0") # If you need to make complex sub-parts in the datasets with configurable options # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig # BUILDER_CONFIG_CLASS = MyBuilderConfig # You will be able to load one or the other configurations in the following list with # data = datasets.load_dataset('my_dataset', 'first_domain') # data = datasets.load_dataset('my_dataset', 'second_domain') BUILDER_CONFIGS = [ SuperEurlexConfig(#version=VERSION, sector=sect, language=lang, feature_version=version, description=SECTOR_DESCRIPTIONS[sect], features=AVAILABLE_FEATURES[sect+("" if version is None else f"_{version}")], citation=_CITATION, url=_HOMEPAGE) for lang in AVAILABLE_LANGUAGES for sect in SECTORS for version in VERSIONS ] #DEFAULT_CONFIG_NAME = "3.DE" # It's not mandatory to have a default configuration. Just use one if it make sense. def _info(self): # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset v = "" if self.config.feature_version=="None" else "_"+self.config.feature_version features = AVAILABLE_FEATURES[self.config.sector+v] info = datasets.DatasetInfo( # This is the description that will appear on the datasets page. description=_DESCRIPTION, # This defines the different columns of the dataset and their types features=features, # Here we define them above because they are different between the two configurations # If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and # specify them. They'll be used if as_supervised=True in builder.as_dataset. # supervised_keys=("sentence", "label"), # Homepage of the dataset for documentation homepage=_HOMEPAGE, # License for the dataset if available license=_LICENSE, # Citation for the dataset citation=_CITATION, ) return info def _split_generators(self, dl_manager): # TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration # If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS # It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files. # By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive urls = {'text': self.config.text_data_url, 'meta': self.config.meta_data_url} try: data_dir = dl_manager.download_and_extract(urls) except FileNotFoundError: raise FileNotFoundError("""The demanded Files weren't found. It could be that the demanded sector isn't yet available in your language of choice""") return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, # These kwargs will be passed to _generate_examples gen_kwargs={ "text": data_dir['text'], "meta": data_dir['meta'], "language": self.config.language, "sector": self.config.sector, 'split': 'train' }, ) ] # method parameters are unpacked from `gen_kwargs` as given in `_split_generators` def _generate_examples(self, text, meta, sector, language, split): # TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset. # The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example. text_data = pd.read_parquet(text) text_data['celex_id'] = text_data['celex_id'].apply(lambda x: str(x.tolist()[0]) if isinstance(x,list) else x) meta_data = pd.read_parquet(meta) meta_data['celex_id'] = meta_data['celex_id'].apply(lambda x: str(x.tolist()[0]) if isinstance(x, np.ndarray) else x) combined_data = pd.merge(text_data, meta_data, on='celex_id') dataset = datasets.Dataset.from_pandas(combined_data) if '__index_level_0__' in dataset.column_names: dataset = dataset.remove_columns('__index_level_0__') for i, sample in enumerate(dataset): yield i, sample if __name__ == '__main__': import datasets as ds import sys print(sys.argv[0]) for sector in SECTORS: for lang in AVAILABLE_LANGUAGES: for version in VERSIONS: v = "" if version is None else f".{version}" print(f'{sector}.{lang}{v}') try: dataset = ds.load_dataset(sys.argv[0],f'{sector}.{lang}{v}') print(dataset) print('\n') except Exception as e: # Handle the exception print("An error occurred: " + str(e.with_traceback(None))) print(f"\n{sector}.{lang} Couldn't be loaded\n")