Sean MacAvaney commited on
Commit
963eb6d
1 Parent(s): 97c0499

commit files to HF hub

Browse files
Files changed (2) hide show
  1. README.md +54 -0
  2. mmarco_v2_fr_train.py +43 -0
README.md ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: '`mmarco/v2/fr/train`'
3
+ viewer: false
4
+ source_datasets: ['irds/mmarco_v2_fr']
5
+ task_categories:
6
+ - text-retrieval
7
+ ---
8
+
9
+ # Dataset Card for `mmarco/v2/fr/train`
10
+
11
+ The `mmarco/v2/fr/train` dataset, provided by the [ir-datasets](https://ir-datasets.com/) package.
12
+ For more information about the dataset, see the [documentation](https://ir-datasets.com/mmarco#mmarco/v2/fr/train).
13
+
14
+ # Data
15
+
16
+ This dataset provides:
17
+ - `queries` (i.e., topics); count=808,731
18
+ - `qrels`: (relevance assessments); count=532,761
19
+ - `docpairs`; count=39,780,811
20
+
21
+ - For `docs`, use [`irds/mmarco_v2_fr`](https://huggingface.co/datasets/irds/mmarco_v2_fr)
22
+
23
+ ## Usage
24
+
25
+ ```python
26
+ from datasets import load_dataset
27
+
28
+ queries = load_dataset('irds/mmarco_v2_fr_train', 'queries')
29
+ for record in queries:
30
+ record # {'query_id': ..., 'text': ...}
31
+
32
+ qrels = load_dataset('irds/mmarco_v2_fr_train', 'qrels')
33
+ for record in qrels:
34
+ record # {'query_id': ..., 'doc_id': ..., 'relevance': ..., 'iteration': ...}
35
+
36
+ docpairs = load_dataset('irds/mmarco_v2_fr_train', 'docpairs')
37
+ for record in docpairs:
38
+ record # {'query_id': ..., 'doc_id_a': ..., 'doc_id_b': ...}
39
+
40
+ ```
41
+
42
+ Note that calling `load_dataset` will download the dataset (or provide access instructions when it's not public) and make a copy of the
43
+ data in 🤗 Dataset format.
44
+
45
+ ## Citation Information
46
+
47
+ ```
48
+ @article{Bonifacio2021MMarco,
49
+ title={{mMARCO}: A Multilingual Version of {MS MARCO} Passage Ranking Dataset},
50
+ author={Luiz Henrique Bonifacio and Israel Campiotti and Roberto Lotufo and Rodrigo Nogueira},
51
+ year={2021},
52
+ journal={arXiv:2108.13897}
53
+ }
54
+ ```
mmarco_v2_fr_train.py ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ """
3
+ """ # TODO
4
+ try:
5
+ import ir_datasets
6
+ except ImportError as e:
7
+ raise ImportError('ir-datasets package missing; `pip install ir-datasets`')
8
+ import datasets
9
+
10
+ IRDS_ID = 'mmarco/v2/fr/train'
11
+ IRDS_ENTITY_TYPES = {'queries': {'query_id': 'string', 'text': 'string'}, 'qrels': {'query_id': 'string', 'doc_id': 'string', 'relevance': 'int64', 'iteration': 'string'}, 'docpairs': {'query_id': 'string', 'doc_id_a': 'string', 'doc_id_b': 'string'}}
12
+
13
+ _CITATION = '@article{Bonifacio2021MMarco,\n title={{mMARCO}: A Multilingual Version of {MS MARCO} Passage Ranking Dataset},\n author={Luiz Henrique Bonifacio and Israel Campiotti and Roberto Lotufo and Rodrigo Nogueira},\n year={2021},\n journal={arXiv:2108.13897}\n}'
14
+
15
+ _DESCRIPTION = "" # TODO
16
+
17
+ class mmarco_v2_fr_train(datasets.GeneratorBasedBuilder):
18
+ BUILDER_CONFIGS = [datasets.BuilderConfig(name=e) for e in IRDS_ENTITY_TYPES]
19
+
20
+ def _info(self):
21
+ return datasets.DatasetInfo(
22
+ description=_DESCRIPTION,
23
+ features=datasets.Features({k: datasets.Value(v) for k, v in IRDS_ENTITY_TYPES[self.config.name].items()}),
24
+ homepage=f"https://ir-datasets.com/mmarco#mmarco/v2/fr/train",
25
+ citation=_CITATION,
26
+ )
27
+
28
+ def _split_generators(self, dl_manager):
29
+ return [datasets.SplitGenerator(name=self.config.name)]
30
+
31
+ def _generate_examples(self):
32
+ dataset = ir_datasets.load(IRDS_ID)
33
+ for i, item in enumerate(getattr(dataset, self.config.name)):
34
+ key = i
35
+ if self.config.name == 'docs':
36
+ key = item.doc_id
37
+ elif self.config.name == 'queries':
38
+ key = item.query_id
39
+ yield key, item._asdict()
40
+
41
+ def as_dataset(self, split=None, *args, **kwargs):
42
+ split = self.config.name # always return split corresponding with this config to avid returning a redundant DatasetDict layer
43
+ return super().as_dataset(split, *args, **kwargs)