Upload 2 files

README.md

@@ -1,5 +1,5 @@
 ---
-pretty_name: HALvest-Raw
+pretty_name: HALvest
 
 configs:
 - config_name: ar
@@ -206,7 +206,7 @@ source_datasets:
 
 
 <div align="center">
-    <h1> HALvest-Raw </h1>
+    <h1> HALvest </h1>
     <h3> Open Scientific Papers Harvested from HAL (Unfiltered) </h3>
 </div>
 
@@ -228,7 +228,7 @@ You can download the dataset using Hugging Face datasets:
 ```py
 from datasets import load_dataset
 
-ds = load_dataset("Madjakul/HALvest-R", "en")
+ds = load_dataset("almanach/HALvest", "en")
 ```
 
 
@@ -337,13 +337,14 @@ The corpus is extracted from the [HAL's open archive](https://hal.science/) whic
 ## Citation
 
 ```bib
-@software{almanach_halvest_2024,
-  author = {Kulumba, Francis and Antoun, Wissam and Vimont, Guillaume and Romary, Laurent},
-  title = {HALvest: Open Scientific Papers Harvested from HAL.},
-  month = April,
-  year = 2024,
-  company = Almanach,
-  url = {https://github.com/Madjakul/HALvesting}
+@misc{kulumba2024harvestingtextualstructureddata,
+      title={Harvesting Textual and Structured Data from the HAL Publication Repository}, 
+      author={Francis Kulumba and Wissam Antoun and Guillaume Vimont and Laurent Romary},
+      year={2024},
+      eprint={2407.20595},
+      archivePrefix={arXiv},
+      primaryClass={cs.DL},
+      url={https://arxiv.org/abs/2407.20595}, 
 }
 ```
 

halvest.py

@@ -0,0 +1,213 @@
+# halvest-r.py
+
+import collections
+import gzip
+import json
+import os
+
+import datasets
+
+logger = datasets.logging.get_logger(__name__)
+
+_DESCRIPTION = "HALvest Raw"
+_URL = "https://huggingface.co/datasets/almanach/HALvest"
+_LICENSE = """
+    The licence terms for HALvest strictly follows those of HAL.
+    Please refer to the below license when using this dataset.
+    - HAL license: https://doc.archives-ouvertes.fr/en/legal-aspects/
+    The corpus is extracted from the HAL's open archive which distributes scientific \
+    publications following open access principles. The corpus is made up of both \
+    creative commons licensed and copyrighted documents (distribution authorized on \
+    HAL by the publisher). This must be considered prior to using this dataset for any \
+    purpose, other than training deep learning models, data mining etc. We do not own \
+    any of the text from which these data has been extracted.
+"""
+_CITATION = """
+    @misc{kulumba2024harvestingtextualstructureddata,
+      title={Harvesting Textual and Structured Data from the HAL Publication Repository}, 
+      author={Francis Kulumba and Wissam Antoun and Guillaume Vimont and Laurent Romary},
+      year={2024},
+      eprint={2407.20595},
+      archivePrefix={arXiv},
+      primaryClass={cs.DL},
+      url={https://arxiv.org/abs/2407.20595}, 
+}
+"""
+_BASE_DATA_PATH = "{language}/"
+_BASE_CHECKSUM_FILENAME = "checksum.sha256"
+
+
+def _languages():
+    """Creates the sorted dictionary of language codes, and language names."""
+    langs = {
+        "Albanian": "sq",
+        "Arabic": "ar",
+        "Armenian": "hy",
+        "Azerbaijani": "az",
+        "Basque": "eu",
+        "Bosnian": "bs",
+        "Breton": "br",
+        "Bulgarian": "bg",
+        "Catalan": "ca",
+        "Chinese": "zh",
+        "Corsican": "co",
+        "Croatian": "hr",
+        "Czech": "cs",
+        "Danish": "da",
+        "English": "en",
+        "Esperanto": "eo",
+        "Estonian": "et",
+        "Filipino": "tl",
+        "Finnish": "fi",
+        "French": "fr",
+        "Galician": "gl",
+        "German": "de",
+        "Greek": "el",
+        "Guarani": "gn",
+        "Hebrew": "he",
+        "Hindi": "hi",
+        "Hungarian": "hu",
+        "Indonesian": "id",
+        "Interlingue": "ie",
+        "Italian": "it",
+        "Japanese": "ja",
+        "Kazakh": "kk",
+        "Korean": "ko",
+        "Lithuanian": "lt",
+        "Macedonian": "mk",
+        "Marathi": "mr",
+        "Norwegian": "no",
+        "Occitan": "oc",
+        "Persian": "fa",
+        "Polish": "pl",
+        "Portuguese": "pt",
+        "Romanian": "ro",
+        "Russian": "ru",
+        "Serbian": "sr",
+        "Slovak": "sk",
+        "Slovenian": "sl",
+        "Spanish": "es",
+        "Swahili": "sw",
+        "Swedish": "sv",
+        "Tamil": "ta",
+        "Tetum": "tet",
+        "Thai": "th",
+        "Tibetan": "bo",
+        "Turkish": "tr",
+        "Turkmen": "tk",
+        "Ukrainian": "uk",
+        "Vietnamese": "vi",
+    }
+    langs = {v: k for k, v in langs.items()}
+    return collections.OrderedDict(sorted(langs.items()))
+
+
+class HALvest_RConfig(datasets.BuilderConfig):
+    """HALvest builder config.
+    Parameters
+    ----------
+    language: str
+        ISO 639 language code.
+    Attributes
+    ----------
+    base_data_path: str
+        f"{self.language}/".
+    """
+
+    def __init__(self, language: str, **kwargs):
+        if language not in _languages():
+            raise ValueError("Invalid language: %s " % language)
+
+        name = f"{language}"
+        description = f"""
+            {_languages()[language]} HALvest dataset from February 2024.
+        """
+        super(HALvest_RConfig, self).__init__(
+            name=name, description=description, **kwargs
+        )
+        self.language = language
+        self.base_data_path = _BASE_DATA_PATH.format(language=language)
+
+
+class HALvest_R(datasets.GeneratorBasedBuilder):
+    """HALvest: Open Scientific Papers Harvested from HAL."""
+
+    BUILDER_CONFIGS = [
+        HALvest_RConfig(language=language, version=datasets.Version("0.1.0"))
+        for language in _languages()
+    ]
+    BUILDER_CONFIG_CLASS = HALvest_RConfig
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "halid": datasets.Value("string"),
+                    "lang": datasets.Value("string"),
+                    "domain": datasets.Sequence("string"),
+                    "timestamp": datasets.Value("string"),
+                    "year": datasets.Value("string"),
+                    "url": datasets.Value("string"),
+                    "text": datasets.Value("string"),
+                    "token_count": datasets.Value("int32"),
+                    "rps_doc_frac_all_caps_words": datasets.Value("float64"),
+                    "rps_doc_frac_lines_end_with_ellipsis": datasets.Value("float64"),
+                    "rps_doc_frac_no_alph_words": datasets.Value("float64"),
+                    "rps_doc_lorem_ipsum": datasets.Value("float64"),
+                    "rps_doc_mean_word_length": datasets.Value("float64"),
+                    "rps_doc_stop_word_fraction": datasets.Value("float64"),
+                    "rps_doc_symbol_to_word_ratio": datasets.Value("float64"),
+                    "rps_doc_frac_unique_words": datasets.Value("float64"),
+                    "rps_doc_unigram_entropy": datasets.Value("float64"),
+                    "rps_doc_word_count": datasets.Value("int64"),
+                    "doc_frac_lines_ending_with_terminal_punctution_mark": datasets.Value("float64"),
+                    "rps_lines_frac_start_with_bulletpoint": datasets.Value("float64"),
+                    "rps_doc_num_sentences": datasets.Value("int64"),
+                    "rps_frac_chars_in_dupe_5grams": datasets.Value("float64"),
+                    "rps_frac_chars_in_dupe_6grams": datasets.Value("float64"),
+                    "rps_frac_chars_in_dupe_7grams": datasets.Value("float64"),
+                    "rps_frac_chars_in_dupe_8grams": datasets.Value("float64"),
+                    "rps_frac_chars_in_dupe_9grams": datasets.Value("float64"),
+                    "rps_frac_chars_in_dupe_10grams": datasets.Value("float64"),
+                    "kenlm_pp": datasets.Value("float64"),
+                }
+            ),
+            supervised_keys=None,
+            homepage=_URL,
+            citation=_CITATION,
+            license=_LICENSE,
+        )
+
+    def _split_generators(self, dl_manager):
+        checksum_path = os.path.join(
+            self.config.base_data_path, _BASE_CHECKSUM_FILENAME
+        )
+        checksum_file = dl_manager.download(checksum_path)
+
+        with open(checksum_file, encoding="utf-8") as f:
+            data_filenames = [line.split("\t")[1] for line in f if line]
+            data_urls = [
+                os.path.join(self.config.base_data_path, data_filename.rstrip("\n"))
+                for data_filename in data_filenames
+            ]
+
+        downloaded_files = dl_manager.download(
+            [url for url in data_urls if url.endswith(".gz")]
+        )
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN, gen_kwargs={"filepaths": downloaded_files}
+            )
+        ]
+
+    def _generate_examples(self, filepaths):
+        id_ = 0
+        for filepath in filepaths:
+            logger.info("Generating examples from = %s", filepath)
+            with gzip.open(open(filepath, "rb"), "rt", encoding="utf-8") as f:
+                for line in f:
+                    js_line = json.loads(line)
+                    yield id_, js_line
+                    id_ += 1