Update README.md
Browse files
README.md
CHANGED
|
@@ -437,3 +437,41 @@ configs:
|
|
| 437 |
- split: dev
|
| 438 |
path: urd/dev-*
|
| 439 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 437 |
- split: dev
|
| 438 |
path: urd/dev-*
|
| 439 |
---
|
| 440 |
+
|
| 441 |
+
# Tatoeba Benchmark (Indian languages only)
|
| 442 |
+
|
| 443 |
+
This benchmark is prepared from the [2023 Tatoeba Challenge](https://github.com/Helsinki-NLP/Tatoeba-Challenge/blob/master/data/README-v2023-09-26.md), by extracting the `dev` and `test` sets for languages spoken in the Indian Republic.
|
| 444 |
+
|
| 445 |
+
The code to download and process the data can be found here in the repo: `data_prep/original_v1/extract.py`
|
| 446 |
+
|
| 447 |
+
Note: This is not the official version of Tatoeba benchmark. Just a processed mirror for Indian languages, made available in HuggingFace for ease of use.
|
| 448 |
+
|
| 449 |
+
## Languages
|
| 450 |
+
|
| 451 |
+
| Language code | Language name |
|
| 452 |
+
|---------------|---------------|
|
| 453 |
+
| asm | Assamese |
|
| 454 |
+
| awa | Awadhi |
|
| 455 |
+
| ben | Bengali |
|
| 456 |
+
| bho | Bhojpuri |
|
| 457 |
+
| brx | Bodo |
|
| 458 |
+
| guj | Gujarati |
|
| 459 |
+
| hin | Hindi |
|
| 460 |
+
| kan | Kannada |
|
| 461 |
+
| kha | Khasi |
|
| 462 |
+
| kok | Konkani |
|
| 463 |
+
| lah | Lahnda |
|
| 464 |
+
| mai | Maithili |
|
| 465 |
+
| mal | Malayalam |
|
| 466 |
+
| mar | Marathi |
|
| 467 |
+
| mni | Manipuri |
|
| 468 |
+
| nep | Nepali |
|
| 469 |
+
| ori | Odia |
|
| 470 |
+
| pan | Panjabi |
|
| 471 |
+
| pli | Pali |
|
| 472 |
+
| san | Sanskrit |
|
| 473 |
+
| sat | Santali |
|
| 474 |
+
| snd | Sindhi |
|
| 475 |
+
| tam | Tamil |
|
| 476 |
+
| tel | Telugu |
|
| 477 |
+
| urd | Urdu |
|