Require data manual download

README.md

@@ -36,13 +36,13 @@ dataset_info:
       dtype: int32
   splits:
   - name: train
-    num_bytes: 5910248
+    num_bytes: 5898752
     num_examples: 4921
   - name: validation
-    num_bytes: 1033253
+    num_bytes: 1031456
     num_examples: 768
-  download_size: 3292236
-  dataset_size: 6943501
+  download_size: 0
+  dataset_size: 6930208
 ---
 
 # Dataset Card for FQuAD
@@ -89,6 +89,16 @@ FQuAD contains 25,000+ question and answer pairs.
 Finetuning CamemBERT on FQuAD yields a F1 score of 88% and an exact match of 77.9%.
 Developped to provide a SQuAD equivalent in the French language. Questions are original and based on high quality Wikipedia articles.
 
+Please, note this dataset is licensed for non-commercial purposes and users must agree to the following terms and conditions:
+1. Use FQuAD only for internal research purposes.
+2. Not make any copy except a safety one.
+3. Not redistribute it (or part of it) in any way, even for free.
+4. Not sell it or use it for any commercial purpose. Contact us for a possible commercial licence.
+5. Mention the corpus origin and Illuin Technology in all publications about experiments using FQuAD.
+6. Redistribute to Illuin Technology any improved or enriched version you could make of that corpus.
+
+Request manually download of the data from: https://fquad.illuin.tech/
+
 ### Supported Tasks and Leaderboards
 
 - `closed-domain-qa`, `text-retrieval`: This dataset is intended to be used for `closed-domain-qa`, but can also be used for information retrieval tasks.

dataset_infos.json

@@ -1 +1 @@
-{"default": {"description": "FQuAD: French Question Answering Dataset\nWe introduce FQuAD, a native French Question Answering Dataset. FQuAD contains 25,000+ question and answer pairs.\nFinetuning CamemBERT on FQuAD yields a F1 score of 88% and an exact match of 77.9%.\n\n", "citation": "@ARTICLE{2020arXiv200206071\n       author = {Martin, d'Hoffschmidt and Maxime, Vidal and\n         Wacim, Belblidia and Tom, Brendl\u00e9},\n        title = \"{FQuAD: French Question Answering Dataset}\",\n      journal = {arXiv e-prints},\n     keywords = {Computer Science - Computation and Language},\n         year = \"2020\",\n        month = \"Feb\",\n          eid = {arXiv:2002.06071},\n        pages = {arXiv:2002.06071},\narchivePrefix = {arXiv},\n       eprint = {2002.06071},\n primaryClass = {cs.CL}\n}\n", "homepage": "https://fquad.illuin.tech/", "license": "", "features": {"context": {"dtype": "string", "id": null, "_type": "Value"}, "questions": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answers": {"feature": {"texts": {"dtype": "string", "id": null, "_type": "Value"}, "answers_starts": {"dtype": "int32", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}}, "supervised_keys": null, "builder_name": "fquad", "config_name": "default", "version": {"version_str": "0.1.0", "description": null, "datasets_version_to_prepare": null, "major": 0, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 5910248, "num_examples": 4921, "dataset_name": "fquad"}, "validation": {"name": "validation", "num_bytes": 1033253, "num_examples": 768, "dataset_name": "fquad"}}, "download_checksums": {"https://storage.googleapis.com/illuin/fquad/train.json.zip": {"num_bytes": 2813123, "checksum": "64f0aea68bacee6ffca7f2f7d56a97e504b2ad2abce057ab6c14768a72c09b47"}, "https://storage.googleapis.com/illuin/fquad/valid.json.zip": {"num_bytes": 479113, "checksum": "53eb7f33573f619f6d56f9e656c6ea6030639ebd663bb445b86b999123be1ef3"}}, "download_size": 3292236, "dataset_size": 6943501, "size_in_bytes": 10235737}}
+{"default": {"description": "FQuAD: French Question Answering Dataset\nWe introduce FQuAD, a native French Question Answering Dataset. FQuAD contains 25,000+ question and answer pairs.\nFinetuning CamemBERT on FQuAD yields a F1 score of 88% and an exact match of 77.9%.\n", "citation": "@ARTICLE{2020arXiv200206071\n       author = {Martin, d'Hoffschmidt and Maxime, Vidal and\n         Wacim, Belblidia and Tom, Brendl\u00e9},\n        title = \"{FQuAD: French Question Answering Dataset}\",\n      journal = {arXiv e-prints},\n     keywords = {Computer Science - Computation and Language},\n         year = \"2020\",\n        month = \"Feb\",\n          eid = {arXiv:2002.06071},\n        pages = {arXiv:2002.06071},\narchivePrefix = {arXiv},\n       eprint = {2002.06071},\n primaryClass = {cs.CL}\n}\n", "homepage": "https://fquad.illuin.tech/", "license": "", "features": {"context": {"dtype": "string", "id": null, "_type": "Value"}, "questions": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answers": {"feature": {"texts": {"dtype": "string", "id": null, "_type": "Value"}, "answers_starts": {"dtype": "int32", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "fquad", "config_name": "default", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 5898752, "num_examples": 4921, "dataset_name": "fquad"}, "validation": {"name": "validation", "num_bytes": 1031456, "num_examples": 768, "dataset_name": "fquad"}}, "download_checksums": {}, "download_size": 0, "post_processing_size": null, "dataset_size": 6930208, "size_in_bytes": 6930208}}

fquad.py

@@ -1,13 +1,21 @@
-"""TODO(fquad): Add a description here."""
+"""FQuAD dataset."""
 
 
 import json
 import os
+from textwrap import dedent
 
 import datasets
 
 
-# TODO(fquad): BibTeX citation
+_HOMEPAGE = "https://fquad.illuin.tech/"
+
+_DESCRIPTION = """\
+FQuAD: French Question Answering Dataset
+We introduce FQuAD, a native French Question Answering Dataset. FQuAD contains 25,000+ question and answer pairs.
+Finetuning CamemBERT on FQuAD yields a F1 score of 88% and an exact match of 77.9%.
+"""
+
 _CITATION = """\
 @ARTICLE{2020arXiv200206071
        author = {Martin, d'Hoffschmidt and Maxime, Vidal and
@@ -25,33 +33,28 @@ archivePrefix = {arXiv},
 }
 """
 
-# TODO(fquad):
-_DESCRIPTION = """\
-FQuAD: French Question Answering Dataset
-We introduce FQuAD, a native French Question Answering Dataset. FQuAD contains 25,000+ question and answer pairs.
-Finetuning CamemBERT on FQuAD yields a F1 score of 88% and an exact match of 77.9%.
 
-"""
+class Fquad(datasets.GeneratorBasedBuilder):
+    """FQuAD dataset."""
 
-_URL = "https://storage.googleapis.com/illuin/fquad/"
-_URLS = {
-    "train": _URL + "train.json.zip",
-    "valid": _URL + "valid.json.zip",
-}
+    VERSION = datasets.Version("1.0.0")
 
+    @property
+    def manual_download_instructions(self):
+        return dedent("""\
+        To access the data for this dataset, you need to request it at:
+        https://fquad.illuin.tech/#download
 
-class Fquad(datasets.GeneratorBasedBuilder):
-    """TODO(fquad): Short description of my dataset."""
+        Unzip the downloaded file with `unzip download-form-fquad1.0.zip -d <path/to/directory>`, into a destination
+        directory <path/to/directory>, which will contain the two data files train.json and valid.json.
 
-    # TODO(fquad): Set up version.
-    VERSION = datasets.Version("0.1.0")
+        To load the dataset, pass the full path to the destination directory
+        in your call to the loading function: `datasets.load_dataset("fquad", data_dir="<path/to/directory>")`
+        """)
 
     def _info(self):
-        # TODO(fquad): Specifies the datasets.DatasetInfo object
         return datasets.DatasetInfo(
-            # This is the description that will appear on the datasets page.
             description=_DESCRIPTION,
-            # datasets.features.FeatureConnectors
             features=datasets.Features(
                 {
                     "context": datasets.Value("string"),
@@ -62,39 +65,28 @@ class Fquad(datasets.GeneratorBasedBuilder):
                     # These are the features of your dataset like images, labels ...
                 }
             ),
-            # If there's a common (input, target) tuple from the features,
-            # specify them here. They'll be used if as_supervised=True in
-            # builder.as_dataset.
-            supervised_keys=None,
-            # Homepage of the dataset for documentation
-            homepage="https://fquad.illuin.tech/",
+            homepage=_HOMEPAGE,
             citation=_CITATION,
         )
 
     def _split_generators(self, dl_manager):
         """Returns SplitGenerators."""
-        # TODO(fquad): Downloads the data and defines the splits
-        # dl_manager is a datasets.download.DownloadManager that can be used to
-        # download and extract URLs
-        download_urls = _URLS
-        dl_dir = dl_manager.download_and_extract(download_urls)
+        data_dir = os.path.abspath(os.path.expanduser(dl_manager.manual_dir))
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 # These kwargs will be passed to _generate_examples
-                gen_kwargs={"filepath": os.path.join(dl_dir["train"], "train.json")},
+                gen_kwargs={"filepath": os.path.join(data_dir, "train.json")},
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.VALIDATION,
                 # These kwargs will be passed to _generate_examples
-                gen_kwargs={"filepath": os.path.join(dl_dir["valid"], "valid.json")},
+                gen_kwargs={"filepath": os.path.join(data_dir, "valid.json")},
             ),
         ]
 
     def _generate_examples(self, filepath):
-
         """Yields examples."""
-        # TODO(fquad): Yields (key, example) tuples from the dataset
         with open(filepath, encoding="utf-8") as f:
             data = json.load(f)
             for id1, examples in enumerate(data["data"]):