ramithuh commited on
Commit
de2ec62
1 Parent(s): 5ab0d37

Include additional details (motivation, citation, provenance etc.)

Browse files
Files changed (1) hide show
  1. README.md +55 -4
README.md CHANGED
@@ -252,12 +252,14 @@ An example of the `Aya Evaluation Suite` looks as follows:
252
 
253
  ## Statistics
254
 
 
 
255
  ### Languages
256
 
257
  <details>
258
  <summary> <b>aya-human-annotated</b> </summary>
259
 
260
- | ISO Code | Language | Resourcedness |
261
  |----------|----------|---------------|
262
  | `tel` | Telugu | Low |
263
  | `yor` | Yorùbá | Low |
@@ -273,7 +275,7 @@ An example of the `Aya Evaluation Suite` looks as follows:
273
  <details>
274
  <summary> <b>dolly-machine-translated</b> </summary>
275
 
276
- | ISO Code | Language | Resourcedness |
277
  |----------|----------|-----------|
278
  | `ace` | Achinese | Low |
279
  | `afr` | Afrikaans | Mid |
@@ -381,7 +383,7 @@ An example of the `Aya Evaluation Suite` looks as follows:
381
  <details>
382
  <summary> <b>dolly-human-edited</b> </summary>
383
 
384
- | ISO Code | Language | Resourcedness |
385
  |----------|----------|-----------|
386
  | `arb` | Arabic | High |
387
  | `fra` | French | High |
@@ -390,4 +392,53 @@ An example of the `Aya Evaluation Suite` looks as follows:
390
  | `spa` | Spanish | High |
391
  | `srp` | Serbian | High |
392
 
393
- </details>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
252
 
253
  ## Statistics
254
 
255
+ The toggled table below lists the breakdown of languages in each subset.
256
+
257
  ### Languages
258
 
259
  <details>
260
  <summary> <b>aya-human-annotated</b> </summary>
261
 
262
+ | ISO Code | Language | Resources |
263
  |----------|----------|---------------|
264
  | `tel` | Telugu | Low |
265
  | `yor` | Yorùbá | Low |
 
275
  <details>
276
  <summary> <b>dolly-machine-translated</b> </summary>
277
 
278
+ | ISO Code | Language | Resources |
279
  |----------|----------|-----------|
280
  | `ace` | Achinese | Low |
281
  | `afr` | Afrikaans | Mid |
 
383
  <details>
384
  <summary> <b>dolly-human-edited</b> </summary>
385
 
386
+ | ISO Code | Language | Resources |
387
  |----------|----------|-----------|
388
  | `arb` | Arabic | High |
389
  | `fra` | French | High |
 
392
  | `spa` | Spanish | High |
393
  | `srp` | Serbian | High |
394
 
395
+ </details>
396
+
397
+ <br>
398
+
399
+ # Motivations & Intentions
400
+
401
+ - **Curation Rationale:** This evaluation suite is tailored to test the generation quality of multilingual models, with the aim of balancing language coverage and human-sourced quality.
402
+ It covers prompts originally written in each language, as well as English-centric translated and manually curated or edited prompts for a linguistically broad but rich testbed.
403
+ The list of languages was established from mT5 and aligned with the annotators’ language list and the NLLB translation model.
404
+
405
+ ---
406
+
407
+ # Additional Information
408
+
409
+ ## Provenance
410
+ - **Methods Used:** combination of original annotations by volunteers, automatic translation, and post-editing of translations by professional annotators.
411
+ - **Methodology Details:**
412
+ - *Source:* Original annotations and translations and post-edits of Dolly
413
+ - *Platform:* [Aya Annotation Platform](https://aya.for.ai/)
414
+ - *Dates of Collection:* Jun 2023 - Dec 2023
415
+
416
+
417
+ ## Dataset Version and Maintenance
418
+ - **Maintenance Status:** Actively Maintained
419
+ - **Version Details:**
420
+ - *Current version:* 1.0
421
+ - *Last Update:* 02/2024
422
+ - *First Release:* 02/2024
423
+ - **Maintenance Plan:** No updates planned.
424
+
425
+
426
+ ## Authorship
427
+
428
+ - **Publishing Organization:** [Cohere For AI](https://cohere.com/research)
429
+ - **Industry Type:** Not-for-profit - Tech
430
+ - **Contact Details:** https://aya.for.ai/
431
+
432
+
433
+ ## Licensing Information
434
+ This dataset can be used for any purpose, whether academic or commercial, under the terms of the [Apache 2.0](https://opensource.org/license/apache-2-0) License.
435
+
436
+
437
+ ## Citation Information
438
+ ```bibtex
439
+ @misc{ayadata2024,
440
+ title = {Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning},
441
+ author = {Shivalika Singh, Freddie Vargus, Daniel D'souza, Börje F. Karlsson, Abinaya Mahendiran, Wei-Yin Ko, Herumb Shandilya, Jay Patel, Deividas Mataciunas, Laura O'Mahony, Mike Zhang, Ramith Hettiarachchi, Joseph Wilson, Marina Machado, Luisa Souza Moura, Dominik Krzemiński, Hakimeh Fadaei, Irem Ergün, Ifeoma Okoh, Aisha Alaagib, Oshan Mudannayake, Zaid Alyafeai, Vu Minh Chien, Sebastian Ruder, Surya Guthikonda, Emad A. Alghamdi, Sebastian Gehrmann, Niklas Muennighoff, Max Bartolo, Julia Kreutzer, Ahmet Üstün, Marzieh Fadaee, Sara Hooker},
442
+ year = 2024,
443
+ }
444
+ ```