Datasets:
Create germandpr-beir.py
Browse files- germandpr-beir.py +104 -0
germandpr-beir.py
ADDED
@@ -0,0 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import datasets
|
2 |
+
|
3 |
+
|
4 |
+
_VERSION = "1.0.0"
|
5 |
+
|
6 |
+
_DESCRIPTION = "Deepset's germanDPR dataset made compatible with BEIR benchmark framework. One version contains " \
|
7 |
+
"the original dataset 1:1 and the other dataset is preprocessed. See official dataset card for " \
|
8 |
+
"usage of dataset with BEIR."
|
9 |
+
|
10 |
+
_SUBSETS = ["queries-original", "corpus-original", "queries-processed", "corpus-processed", "qrels"]
|
11 |
+
|
12 |
+
|
13 |
+
class GermanDPRBeir(datasets.GeneratorBasedBuilder):
|
14 |
+
BUILDER_CONFIGS = (
|
15 |
+
[
|
16 |
+
datasets.BuilderConfig(
|
17 |
+
name="queries-original",
|
18 |
+
description=f"BEIR queries created 1:1 from deepset/germanDPR.",
|
19 |
+
version=_VERSION,
|
20 |
+
),
|
21 |
+
datasets.BuilderConfig(
|
22 |
+
name="corpus-original",
|
23 |
+
description=f"BEIR corpus created 1:1 from deepset/germanDPR.",
|
24 |
+
version=_VERSION,
|
25 |
+
),
|
26 |
+
datasets.BuilderConfig(
|
27 |
+
name="queries-processed",
|
28 |
+
description=f"BEIR queries created and further text-processed from deepset/germanDPR.",
|
29 |
+
version=_VERSION,
|
30 |
+
),
|
31 |
+
datasets.BuilderConfig(
|
32 |
+
name="corpus-processed",
|
33 |
+
description=f"BEIR corpus created and further text-processed from deepset/germanDPR.",
|
34 |
+
version=_VERSION,
|
35 |
+
),
|
36 |
+
datasets.BuilderConfig(
|
37 |
+
name="qrels",
|
38 |
+
description=f"BEIR qrels created from deepset/germanDPR for train and test split.",
|
39 |
+
version=_VERSION,
|
40 |
+
)
|
41 |
+
]
|
42 |
+
)
|
43 |
+
|
44 |
+
DEFAULT_CONFIG_NAME = "qrels"
|
45 |
+
|
46 |
+
def _info(self):
|
47 |
+
name = self.config.name
|
48 |
+
_SPLITS = ["queries-original", "corpus-original", "queries-processed", "corpus-processed", "qrels"]
|
49 |
+
|
50 |
+
if name.startswith("queries"):
|
51 |
+
features = {
|
52 |
+
"_id": datasets.Value("string"),
|
53 |
+
"text": datasets.Value("string")
|
54 |
+
}
|
55 |
+
elif name.startswith("corpus"):
|
56 |
+
features = {
|
57 |
+
"_id": datasets.Value("string"),
|
58 |
+
"title": datasets.Value("string"),
|
59 |
+
"text": datasets.Value("string"),
|
60 |
+
}
|
61 |
+
else:
|
62 |
+
# name == qrels
|
63 |
+
features = {
|
64 |
+
"query-id": datasets.Value("string"),
|
65 |
+
"corpus-id": datasets.Value("string"),
|
66 |
+
"score": datasets.Value("int32")
|
67 |
+
}
|
68 |
+
|
69 |
+
return datasets.DatasetInfo(
|
70 |
+
description=f"{_DESCRIPTION}\n{self.config.description}",
|
71 |
+
features=datasets.Features(features),
|
72 |
+
supervised_keys=None,
|
73 |
+
homepage="https://huggingface.co/datasets/PM-AI/germandpr-beir",
|
74 |
+
)
|
75 |
+
|
76 |
+
def _split_generators(self, dl_manager):
|
77 |
+
"""Returns SplitGenerators."""
|
78 |
+
_SPLITS = ["queries-original", "corpus-original", "queries-processed", "corpus-processed", "qrels"]
|
79 |
+
|
80 |
+
name = self.config.name
|
81 |
+
if name == "qrels":
|
82 |
+
dl_path = dl_manager.download([
|
83 |
+
"https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/qrels/test.tsv",
|
84 |
+
"https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/qrels/train.tsv"
|
85 |
+
])
|
86 |
+
else:
|
87 |
+
dl_path = dl_manager.download(f"https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/{name}.jsonl")
|
88 |
+
|
89 |
+
return [
|
90 |
+
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": dl_path, "split": "train"}),
|
91 |
+
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": dl_path, "split": "test"})
|
92 |
+
]
|
93 |
+
|
94 |
+
def _generate_examples(self, filepath, split):
|
95 |
+
"""Yields examples."""
|
96 |
+
name = self.config.name
|
97 |
+
if name.startswith("queries"):
|
98 |
+
yield 0, {"_id": "1", "text": "text"}
|
99 |
+
elif name.startswith("corpus"):
|
100 |
+
yield 0, {"_id": "1", "title": "title", "text": "text"}
|
101 |
+
else:
|
102 |
+
# name == qrels
|
103 |
+
filepath = [x for x in filepath if x.endswith(f"{split}.tsv")]
|
104 |
+
yield 0, {"query-id": "", "corpus-id": "", "score": 1}
|