Datasets:
mteb
/

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
loicmagne commited on
Commit
7e95551
1 Parent(s): 16f46f0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +271 -1
README.md CHANGED
@@ -23,9 +23,279 @@ language:
23
  - ta
24
  - te
25
  - ur
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26
  configs:
27
  - config_name: default
28
  data_files:
29
  - split: test
30
  path: test.parquet
31
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
  - ta
24
  - te
25
  - ur
26
+ language_details: >-
27
+ asm_Beng, ben_Beng, brx_Deva, doi_Deva, eng_Latn, gom_Deva, guj_Gujr,
28
+ hin_Deva, kan_Knda, kas_Arab, mai_Deva, mal_Mlym, mar_Deva, mni_Mtei,
29
+ npi_Deva, ory_Orya, pan_Guru, san_Deva, sat_Olck, snd_Deva, tam_Taml,
30
+ tel_Telu, urd_Arab
31
+ license: cc-by-4.0
32
+ language_creators:
33
+ - expert-generated
34
+ multilinguality:
35
+ - multilingual
36
+ - translation
37
+ pretty_name: in22-conv
38
+ size_categories:
39
+ - 1K<n<10K
40
+ task_categories:
41
+ - translation
42
  configs:
43
  - config_name: default
44
  data_files:
45
  - split: test
46
  path: test.parquet
47
+ ---
48
+
49
+ # IN22-Conv
50
+
51
+ IN-22 is a newly created comprehensive benchmark for evaluating machine translation performance in multi-domain, n-way parallel contexts across 22 Indic languages. IN22-Conv is the conversation domain subset of IN22. It is designed to assess translation quality in typical day-to-day conversational-style applications. The evaluation subset consists of 1503 sentences translated across 22 Indic languages enabling evaluation of MT systems across 506 directions.
52
+
53
+ Currently, we use it for sentence-level evaluation of MT systems but it can be repurposed for document translation evaluation as well.
54
+
55
+ Here is the domain distribution of our IN22-Conv evaluation subset.
56
+
57
+ <table style="width:25%">
58
+ <tr>
59
+ <td>domain</td>
60
+ <td>count</td>
61
+ </tr>
62
+ <tr>
63
+ <td>hobbies</td>
64
+ <td>120</td>
65
+ </tr>
66
+ <tr>
67
+ <td>daily_dialogue</td>
68
+ <td>117</td>
69
+ </tr>
70
+ <tr>
71
+ <td>government</td>
72
+ <td>116</td>
73
+ </tr>
74
+ <tr>
75
+ <td>geography</td>
76
+ <td>114</td>
77
+ </tr>
78
+ <tr>
79
+ <td>sports</td>
80
+ <td>100</td>
81
+ </tr>
82
+ <tr>
83
+ <td>entertainment</td>
84
+ <td>97</td>
85
+ </tr>
86
+ <tr>
87
+ <td>history</td>
88
+ <td>97</td>
89
+ </tr>
90
+ <tr>
91
+ <td>legal</td>
92
+ <td>96</td>
93
+ </tr>
94
+ <tr>
95
+ <td>arts</td>
96
+ <td>95</td>
97
+ </tr>
98
+ <tr>
99
+ <td>college_life</td>
100
+ <td>94</td>
101
+ </tr>
102
+ <tr>
103
+ <td>tourism</td>
104
+ <td>91</td>
105
+ </tr>
106
+ <tr>
107
+ <td>school_life</td>
108
+ <td>87</td>
109
+ </tr>
110
+ <tr>
111
+ <td>insurance</td>
112
+ <td>82</td>
113
+ </tr>
114
+ <tr>
115
+ <td>culture</td>
116
+ <td>73</td>
117
+ </tr>
118
+ <tr>
119
+ <td>healthcare</td>
120
+ <td>67</td>
121
+ </tr>
122
+ <tr>
123
+ <td>banking</td>
124
+ <td>57</td>
125
+ </tr>
126
+ <tr>
127
+ <td>total</td>
128
+ <td>1503</td>
129
+ </tr>
130
+ </table>
131
+
132
+ Please refer to the `Appendix E: Dataset Card` of the [preprint](https://arxiv.org/abs/2305.16307) on detailed description of dataset curation, annotation and quality control process.
133
+
134
+
135
+ ### Dataset Structure
136
+
137
+ #### Dataset Fields
138
+
139
+ - `id`: Row number for the data entry, starting at 1.
140
+ - `doc_id`: Unique identifier of the conversation.
141
+ - `sent_id`: Unique identifier of the sentence order in each conversation.
142
+ - `topic`: The specific topic of the conversation within the domain.
143
+ - `domain`: The domain of the conversation.
144
+ - `prompt`: The prompt provided to annotators to simulate the conversation.
145
+ - `scenario`: The scenario or context in which the conversation takes place.
146
+ - `speaker`: The speaker identifier in the conversation.
147
+ - `turn`: The turn within the conversation.
148
+
149
+ #### Data Instances
150
+
151
+ A sample from the `gen` split for the English language (`eng_Latn` config) is provided below. All configurations have the same structure, and all sentences are aligned across configurations and splits.
152
+
153
+ ```python
154
+ {
155
+ "id": 1,
156
+ "doc_id": 0,
157
+ "sent_id": 1,
158
+ "topic": "Festivities",
159
+ "domain": "culture",
160
+ "prompt": "14th April a holiday",
161
+ "scenario": "Historical importance",
162
+ "speaker": 1,
163
+ "turn": 1,
164
+ "sentence": "Mom, let's go for a movie tomorrow."
165
+ }
166
+ ```
167
+
168
+ When using a hyphenated pairing or using the `all` function, data will be presented as follows:
169
+
170
+ ```python
171
+ {
172
+ "id": 1,
173
+ "doc_id": 0,
174
+ "sent_id": 1,
175
+ "topic": "Festivities",
176
+ "domain": "culture",
177
+ "prompt": "14th April a holiday",
178
+ "scenario": "Historical importance",
179
+ "speaker": 1,
180
+ "turn": 1,
181
+ "sentence_eng_Latn": "Mom, let's go for a movie tomorrow.",
182
+ "sentence_hin_Deva": "माँ, चलो कल एक फिल्म देखने चलते हैं।"
183
+ }
184
+ ```
185
+
186
+ #### Sample Conversation
187
+
188
+ <table>
189
+ <tr>
190
+ <td>Speaker</td>
191
+ <td>Turn</td>
192
+ </tr>
193
+ <tr>
194
+ <td>Speaker 1</td>
195
+ <td>Mom, let&#39;s go for a movie tomorrow. I don&#39;t have to go to school. It is a holiday.</td>
196
+ </tr>
197
+ <tr>
198
+ <td>Speaker 2</td>
199
+ <td>Oh, tomorrow is the 14th of April right? Your dad will also have the day off from work. We can make a movie plan!</td>
200
+ </tr>
201
+ <tr>
202
+ <td>Speaker 1</td>
203
+ <td>That&#39;s a good news! Why is it a holiday though? Are all schools, colleges and offices closed tomorrow?</td>
204
+ </tr>
205
+ <tr>
206
+ <td>Speaker 2</td>
207
+ <td>It is Ambedkar Jayanti tomorrow! This day is celebrated annually to mark the birth of Dr. B. R Ambedkar. Have you heard of him?</td>
208
+ </tr>
209
+ <tr>
210
+ <td>Speaker 1</td>
211
+ <td>I think I have seen him in my History and Civics book. Is he related to our Constitution?</td>
212
+ </tr>
213
+ <tr>
214
+ <td>Speaker 2</td>
215
+ <td>Absolutely! He is known as the father of the Indian Constitution. He was a civil rights activist who played a major role in formulating the Constitution. He played a crucial part in shaping the vibrant democratic structure that India prides itself upon.</td>
216
+ </tr>
217
+ <tr>
218
+ <td></td>
219
+ <td>...</td>
220
+ </tr>
221
+ </table>
222
+
223
+
224
+ ### Usage Instructions
225
+
226
+ ```python
227
+ from datasets import load_dataset
228
+
229
+ # download and load all the pairs
230
+ dataset = load_dataset("ai4bharat/IN22-Conv", "all")
231
+
232
+ # download and load specific pairs
233
+ dataset = load_dataset("ai4bharat/IN22-Conv", "eng_Latn-hin_Deva")
234
+ ```
235
+
236
+ ### Languages Covered
237
+
238
+ <table style="width: 40%">
239
+ <tr>
240
+ <td>Assamese (asm_Beng)</td>
241
+ <td>Kashmiri (Arabic) (kas_Arab)</td>
242
+ <td>Punjabi (pan_Guru)</td>
243
+ </tr>
244
+ <tr>
245
+ <td>Bengali (ben_Beng)</td>
246
+ <td>Kashmiri (Devanagari) (kas_Deva)</td>
247
+ <td>Sanskrit (san_Deva)</td>
248
+ </tr>
249
+ <tr>
250
+ <td>Bodo (brx_Deva)</td>
251
+ <td>Maithili (mai_Deva)</td>
252
+ <td>Santali (sat_Olck)</td>
253
+ </tr>
254
+ <tr>
255
+ <td>Dogri (doi_Deva)</td>
256
+ <td>Malayalam (mal_Mlym)</td>
257
+ <td>Sindhi (Arabic) (snd_Arab)</td>
258
+ </tr>
259
+ <tr>
260
+ <td>English (eng_Latn)</td>
261
+ <td>Marathi (mar_Deva)</td>
262
+ <td>Sindhi (Devanagari) (snd_Deva)</td>
263
+ </tr>
264
+ <tr>
265
+ <td>Konkani (gom_Deva)</td>
266
+ <td>Manipuri (Bengali) (mni_Beng)</td>
267
+ <td>Tamil (tam_Taml)</td>
268
+ </tr>
269
+ <tr>
270
+ <td>Gujarati (guj_Gujr)</td>
271
+ <td>Manipuri (Meitei) (mni_Mtei)</td>
272
+ <td>Telugu (tel_Telu)</td>
273
+ </tr>
274
+ <tr>
275
+ <td>Hindi (hin_Deva)</td>
276
+ <td>Nepali (npi_Deva)</td>
277
+ <td>Urdu (urd_Arab)</td>
278
+ </tr>
279
+ <tr>
280
+ <td>Kannada (kan_Knda)</td>
281
+ <td>Odia (ory_Orya)</td>
282
+ </tr>
283
+ </table>
284
+
285
+
286
+ ### Citation
287
+
288
+ If you consider using our work then please cite using:
289
+
290
+ ```
291
+ @article{gala2023indictrans,
292
+ title={IndicTrans2: Towards High-Quality and Accessible Machine Translation Models for all 22 Scheduled Indian Languages},
293
+ author={Jay Gala and Pranjal A Chitale and A K Raghavan and Varun Gumma and Sumanth Doddapaneni and Aswanth Kumar M and Janki Atul Nawale and Anupama Sujatha and Ratish Puduppully and Vivek Raghavan and Pratyush Kumar and Mitesh M Khapra and Raj Dabre and Anoop Kunchukuttan},
294
+ journal={Transactions on Machine Learning Research},
295
+ issn={2835-8856},
296
+ year={2023},
297
+ url={https://openreview.net/forum?id=vfT4YuzAYA},
298
+ note={}
299
+ }
300
+ ```
301
+