vjeronymo2
commited on
Commit
•
4187a7f
1
Parent(s):
80abb16
Adding readme and deleting script comments
Browse files- README.md +61 -0
- mrobust.py +1 -142
README.md
ADDED
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Dataset Summary
|
2 |
+
|
3 |
+
**mRobust** is a multilingual version of the [TREC 2004 Robust passage ranking dataset](https://trec.nist.gov/data/robust/04.guidelines.html).
|
4 |
+
For more information, checkout our papers:
|
5 |
+
<!-- * [**mRobust: A Multilingual Version of the MS MARCO Passage Ranking Dataset**](https://arxiv.org/abs/2108.13897)
|
6 |
+
* [**A cost-benefit analysis of cross-lingual transfer methods**](https://arxiv.org/abs/2105.06813) -->
|
7 |
+
|
8 |
+
|
9 |
+
The current version is composed 10 languages: Chinese, French, German, Indonesian, Italian, Portuguese, Russian, Spanish, Dutch and Vietnamese.
|
10 |
+
|
11 |
+
|
12 |
+
### Supported languages
|
13 |
+
|
14 |
+
| Language name | Language code |
|
15 |
+
|---------------|---------------|
|
16 |
+
| English | english |
|
17 |
+
| Chinese | chinese |
|
18 |
+
| French | french |
|
19 |
+
| German | german |
|
20 |
+
| Indonesian | indonesian |
|
21 |
+
| Italian | italian |
|
22 |
+
| Portuguese | portuguese |
|
23 |
+
| Russian | russian |
|
24 |
+
| Spanish | spanish |
|
25 |
+
| Dutch | dutch |
|
26 |
+
| Vietnamese | vietnamese |
|
27 |
+
|
28 |
+
|
29 |
+
# Dataset Structure
|
30 |
+
|
31 |
+
You can load mRobust dataset by choosing a specific language. We include the translated collections of documents and queries.
|
32 |
+
|
33 |
+
#### Queries
|
34 |
+
|
35 |
+
```python
|
36 |
+
>>> dataset = load_dataset('unicamp-dl/mrobust', 'queries-spanish')
|
37 |
+
>>> dataset['queries'][1]
|
38 |
+
{'id': '302', 'text': '¿Está controlada la enfermedad de la poliomielitis (polio) en el mundo?'}
|
39 |
+
```
|
40 |
+
|
41 |
+
#### Collection
|
42 |
+
|
43 |
+
```python
|
44 |
+
>>> dataset = load_dataset('unicamp-dl/mrobust', 'collection-portuguese')
|
45 |
+
>>> dataset['collection'][5]
|
46 |
+
{'id': 'FT931-16660', 'text': '930105 FT 05 JAN 93 / Cenelec: Correção O endereço do Cenelec, Comitê Europeu de Normalização Eletrotécnica, estava incorreto na edição de ontem. É Rue de Stassart 35, B-1050, Bruxelas, Tel (322) 519 6871. CEN, Comitê Europeu de Normalização, está localizado na Rue de Stassart 36, B-1050, Bruxelas, Tel 519 6811.'}
|
47 |
+
```
|
48 |
+
|
49 |
+
|
50 |
+
# Citation Information
|
51 |
+
Soon.
|
52 |
+
<!-- ```
|
53 |
+
@misc{bonifacio2021mmarco,
|
54 |
+
title={mRobust: A Multilingual Version of MS MARCO Passage Ranking Dataset},
|
55 |
+
author={Luiz Henrique Bonifacio and Vitor Jeronymo and Hugo Queiroz Abonizio and Israel Campiotti and Marzieh Fadaee and and Roberto Lotufo and Rodrigo Nogueira},
|
56 |
+
year={2021},
|
57 |
+
eprint={2108.13897},
|
58 |
+
archivePrefix={arXiv},
|
59 |
+
primaryClass={cs.CL}
|
60 |
+
}
|
61 |
+
``` -->
|
mrobust.py
CHANGED
@@ -36,31 +36,6 @@ _DESCRIPTION = """
|
|
36 |
Robust04 translated datasets
|
37 |
"""
|
38 |
|
39 |
-
|
40 |
-
# def generate_examples_triples(filepath, collection_path, queries_path):
|
41 |
-
# collection = {}
|
42 |
-
# with open(collection_path, encoding="utf-8") as f:
|
43 |
-
# for line in f:
|
44 |
-
# doc_id, doc = line.rstrip().split("\t")
|
45 |
-
# collection[doc_id] = doc
|
46 |
-
|
47 |
-
# queries = {}
|
48 |
-
# with open(queries_path, encoding="utf-8") as f:
|
49 |
-
# for line in f:
|
50 |
-
# query_id, query = line.rstrip().split("\t")
|
51 |
-
# queries[query_id] = query
|
52 |
-
|
53 |
-
# with open(filepath, encoding="utf-8") as f:
|
54 |
-
# for (idx, line) in enumerate(f):
|
55 |
-
# query_id, pos_id, neg_id = line.rstrip().split("\t")
|
56 |
-
# features = {
|
57 |
-
# "query": queries[query_id],
|
58 |
-
# "positive": collection[pos_id],
|
59 |
-
# "negative": collection[neg_id],
|
60 |
-
# }
|
61 |
-
# yield idx, features
|
62 |
-
|
63 |
-
|
64 |
def generate_examples_tuples(filepath):
|
65 |
with open(filepath, encoding="utf-8") as f:
|
66 |
for (idx, line) in enumerate(f):
|
@@ -71,48 +46,9 @@ def generate_examples_tuples(filepath):
|
|
71 |
}
|
72 |
yield idx, features
|
73 |
|
74 |
-
|
75 |
-
# def generate_examples_runs(filepath, collection_path, queries_path):
|
76 |
-
# collection = {}
|
77 |
-
# with open(collection_path, encoding="utf-8") as f:
|
78 |
-
# for line in f:
|
79 |
-
# doc_id, doc = line.rstrip().split("\t")
|
80 |
-
# collection[doc_id] = doc
|
81 |
-
|
82 |
-
# queries = {}
|
83 |
-
# with open(queries_path, encoding="utf-8") as f:
|
84 |
-
# for line in f:
|
85 |
-
# query_id, query = line.rstrip().split("\t")
|
86 |
-
# queries[query_id] = query
|
87 |
-
|
88 |
-
# qid_to_ranked_candidate_passages = {}
|
89 |
-
# with open(filepath, encoding="utf-8") as f:
|
90 |
-
# for line in f:
|
91 |
-
# qid, pid, rank = line.rstrip().split("\t")
|
92 |
-
# if qid not in qid_to_ranked_candidate_passages:
|
93 |
-
# qid_to_ranked_candidate_passages[qid] = []
|
94 |
-
# qid_to_ranked_candidate_passages[qid].append(pid)
|
95 |
-
|
96 |
-
# for (idx, qid) in enumerate(qid_to_ranked_candidate_passages):
|
97 |
-
# features = {
|
98 |
-
# "id": qid,
|
99 |
-
# "query": queries[qid],
|
100 |
-
# "passages": [
|
101 |
-
# {
|
102 |
-
# "id": pid,
|
103 |
-
# "passage": collection[pid],
|
104 |
-
# }
|
105 |
-
# for pid in qid_to_ranked_candidate_passages[qid]
|
106 |
-
# ],
|
107 |
-
# }
|
108 |
-
# yield idx, features
|
109 |
-
|
110 |
-
|
111 |
_BASE_URLS = {
|
112 |
"collections": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/collections/",
|
113 |
"queries": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/queries/",
|
114 |
-
# "runs": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/runs/",
|
115 |
-
# "train": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/triples.train.ids.small.tsv",
|
116 |
}
|
117 |
|
118 |
LANGUAGES = [
|
@@ -133,14 +69,6 @@ LANGUAGES = [
|
|
133 |
class MRobust(datasets.GeneratorBasedBuilder):
|
134 |
|
135 |
BUILDER_CONFIGS = (
|
136 |
-
# [
|
137 |
-
# datasets.BuilderConfig(
|
138 |
-
# name=language,
|
139 |
-
# description=f"{language.capitalize()} triples",
|
140 |
-
# version=datasets.Version("1.0.0"),
|
141 |
-
# )
|
142 |
-
# for language in LANGUAGES
|
143 |
-
# ]
|
144 |
[
|
145 |
datasets.BuilderConfig(
|
146 |
name=f"collection-{language}",
|
@@ -157,14 +85,6 @@ class MRobust(datasets.GeneratorBasedBuilder):
|
|
157 |
)
|
158 |
for language in LANGUAGES
|
159 |
]
|
160 |
-
# + [
|
161 |
-
# datasets.BuilderConfig(
|
162 |
-
# name=f"runs-{language}",
|
163 |
-
# description=f"{language.capitalize()} runs",
|
164 |
-
# version=datasets.Version("1.0.0"),
|
165 |
-
# )
|
166 |
-
# for language in LANGUAGES
|
167 |
-
# ]
|
168 |
)
|
169 |
|
170 |
DEFAULT_CONFIG_NAME = "english"
|
@@ -176,23 +96,6 @@ class MRobust(datasets.GeneratorBasedBuilder):
|
|
176 |
"id": datasets.Value("string"),
|
177 |
"text": datasets.Value("string"),
|
178 |
}
|
179 |
-
# elif name.startswith("runs"):
|
180 |
-
# features = {
|
181 |
-
# "id": datasets.Value("int32"),
|
182 |
-
# "query": datasets.Value("string"),
|
183 |
-
# "passages": datasets.Sequence(
|
184 |
-
# {
|
185 |
-
# "id": datasets.Value("int32"),
|
186 |
-
# "passage": datasets.Value("string"),
|
187 |
-
# }
|
188 |
-
# ),
|
189 |
-
# }
|
190 |
-
# else:
|
191 |
-
# features = {
|
192 |
-
# "query": datasets.Value("string"),
|
193 |
-
# "positive": datasets.Value("string"),
|
194 |
-
# "negative": datasets.Value("string"),
|
195 |
-
# }
|
196 |
|
197 |
return datasets.DatasetInfo(
|
198 |
description=f"{_DESCRIPTION}\n{self.config.description}",
|
@@ -213,53 +116,9 @@ class MRobust(datasets.GeneratorBasedBuilder):
|
|
213 |
url = _BASE_URLS["queries"] + self.config.name[8:] + "_queries.tsv"
|
214 |
dl_path = dl_manager.download_and_extract(url)
|
215 |
return (datasets.SplitGenerator(name="queries", gen_kwargs={"filepath": dl_path}),)
|
216 |
-
# elif self.config.name.startswith("runs"):
|
217 |
-
# urls = {
|
218 |
-
# "collection": _BASE_URLS["collections"] + self.config.name[5:] + "_collection.tsv",
|
219 |
-
# "queries": _BASE_URLS["queries-dev"] + self.config.name[5:] + "_queries.dev.tsv",
|
220 |
-
# "run": _BASE_URLS["runs"] + "run.bm25_" + self.config.name[5:] + ".txt",
|
221 |
-
# }
|
222 |
-
|
223 |
-
# dl_path = dl_manager.download_and_extract(urls)
|
224 |
-
# return (
|
225 |
-
# datasets.SplitGenerator(
|
226 |
-
# name="bm25",
|
227 |
-
# gen_kwargs={
|
228 |
-
# "filepath": dl_path["run"],
|
229 |
-
# "args": {
|
230 |
-
# "collection": dl_path["collection"],
|
231 |
-
# "queries": dl_path["queries"],
|
232 |
-
# },
|
233 |
-
# },
|
234 |
-
# ),
|
235 |
-
# )
|
236 |
-
# else:
|
237 |
-
# urls = {
|
238 |
-
# "collection": _BASE_URLS["collections"] + self.config.name + "_collection.tsv",
|
239 |
-
# "queries": _BASE_URLS["queries-train"] + self.config.name + "_queries.train.tsv",
|
240 |
-
# "train": _BASE_URLS["train"],
|
241 |
-
# }
|
242 |
-
# dl_path = dl_manager.download_and_extract(urls)
|
243 |
-
|
244 |
-
# return [
|
245 |
-
# datasets.SplitGenerator(
|
246 |
-
# name=datasets.Split.TRAIN,
|
247 |
-
# gen_kwargs={
|
248 |
-
# "filepath": dl_path["train"],
|
249 |
-
# "args": {
|
250 |
-
# "collection": dl_path["collection"],
|
251 |
-
# "queries": dl_path["queries"],
|
252 |
-
# },
|
253 |
-
# },
|
254 |
-
# )
|
255 |
-
# ]
|
256 |
|
257 |
def _generate_examples(self, filepath, args=None):
|
258 |
"""Yields examples."""
|
259 |
|
260 |
if self.config.name.startswith("collection") or self.config.name.startswith("queries"):
|
261 |
-
return generate_examples_tuples(filepath)
|
262 |
-
# if self.config.name.startswith("runs"):
|
263 |
-
# return generate_examples_runs(filepath, args["collection"], args["queries"])
|
264 |
-
# else:
|
265 |
-
# return generate_examples_triples(filepath, args["collection"], args["queries"])
|
|
|
36 |
Robust04 translated datasets
|
37 |
"""
|
38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
def generate_examples_tuples(filepath):
|
40 |
with open(filepath, encoding="utf-8") as f:
|
41 |
for (idx, line) in enumerate(f):
|
|
|
46 |
}
|
47 |
yield idx, features
|
48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
_BASE_URLS = {
|
50 |
"collections": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/collections/",
|
51 |
"queries": "https://huggingface.co/datasets/unicamp-dl/mrobust/resolve/main/data/queries/",
|
|
|
|
|
52 |
}
|
53 |
|
54 |
LANGUAGES = [
|
|
|
69 |
class MRobust(datasets.GeneratorBasedBuilder):
|
70 |
|
71 |
BUILDER_CONFIGS = (
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
[
|
73 |
datasets.BuilderConfig(
|
74 |
name=f"collection-{language}",
|
|
|
85 |
)
|
86 |
for language in LANGUAGES
|
87 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
)
|
89 |
|
90 |
DEFAULT_CONFIG_NAME = "english"
|
|
|
96 |
"id": datasets.Value("string"),
|
97 |
"text": datasets.Value("string"),
|
98 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
return datasets.DatasetInfo(
|
101 |
description=f"{_DESCRIPTION}\n{self.config.description}",
|
|
|
116 |
url = _BASE_URLS["queries"] + self.config.name[8:] + "_queries.tsv"
|
117 |
dl_path = dl_manager.download_and_extract(url)
|
118 |
return (datasets.SplitGenerator(name="queries", gen_kwargs={"filepath": dl_path}),)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
|
120 |
def _generate_examples(self, filepath, args=None):
|
121 |
"""Yields examples."""
|
122 |
|
123 |
if self.config.name.startswith("collection") or self.config.name.startswith("queries"):
|
124 |
+
return generate_examples_tuples(filepath)
|
|
|
|
|
|
|
|