ramithuh commited on
Commit
8a622ea
1 Parent(s): 8de11f1

Add remarks about dolly subsets

Browse files
Files changed (1) hide show
  1. README.md +45 -4
README.md CHANGED
@@ -196,9 +196,9 @@ configs:
196
 
197
  `Aya Evaluation Suite` contains a total of 25,750 open-ended conversation-style prompts to evaluate multilingual open-ended generation quality.\
198
  To strike a balance between language coverage and the quality that comes with human curation, we create an evaluation suite that includes:
199
- 1) human-curated examples in 7 languages (`tur,eng,yor,arb,zho,por,tel`).
200
- 2) machine-translations of handpicked examples into 101 languages.
201
- 3) human-post-edited translations into 6 languages (`hin,srp,rus,fra,arb,spa`).
202
 
203
  ---
204
 
@@ -239,11 +239,15 @@ aya_eval = load_dataset("CohereForAI/aya_evaluation_suite", "dataset")
239
  - `targets`: Completion or output of the language model. (Not applicable for `dolly-human-edited`)
240
  - `language`: The language of the `prompt` and `completion.`
241
  - `script`: The writing system of the language.
 
242
 
243
  ## Data Instances
244
 
245
- An example of the `Aya Evaluation Suite` looks as follows:
246
 
 
 
 
247
  ```json
248
  {
249
  "id": 42,
@@ -253,6 +257,43 @@ An example of the `Aya Evaluation Suite` looks as follows:
253
  "script": "Latn",
254
  }
255
  ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
256
 
257
  ## Statistics
258
 
 
196
 
197
  `Aya Evaluation Suite` contains a total of 25,750 open-ended conversation-style prompts to evaluate multilingual open-ended generation quality.\
198
  To strike a balance between language coverage and the quality that comes with human curation, we create an evaluation suite that includes:
199
+ 1) human-curated examples in 7 languages (`tur,eng,yor,arb,zho,por,tel`) → `aya-human-annotated`.
200
+ 2) machine-translations of handpicked examples into 101 languages → `dolly-machine-translated`.
201
+ 3) human-post-edited translations into 6 languages (`hin,srp,rus,fra,arb,spa`) → `dolly-human-edited`.
202
 
203
  ---
204
 
 
239
  - `targets`: Completion or output of the language model. (Not applicable for `dolly-human-edited`)
240
  - `language`: The language of the `prompt` and `completion.`
241
  - `script`: The writing system of the language.
242
+ - `source_id`: Corresponding original row index from the [databricks-dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k) dataset (Field applicable only for subsets `dolly-machine-translated` & `dolly-human-edited`)
243
 
244
  ## Data Instances
245
 
246
+ Example data instances from the `Aya Evaluation Suite` subsets are listed in the toggled sections below.
247
 
248
+ <details>
249
+ <summary> <b>aya-human-annotated</b> </summary>
250
+
251
  ```json
252
  {
253
  "id": 42,
 
257
  "script": "Latn",
258
  }
259
  ```
260
+ </details>
261
+
262
+
263
+ <b>Dolly-machine-translated and dolly-human-edited</b>
264
+
265
+ - These two subsets are parallel datasets (data instances can be mapped using the `id`).
266
+ - Note that in the `dolly-machine-translated` subset, we also include the original English subset (`id 1-200`) that is translated into 101 languages. Furthermore, the field `id` can be used to filter out the translation of the same data instance across the languages.
267
+ - The `source_id` field contains the corresponding original row index from the [databricks-dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k) dataset.
268
+ <details>
269
+ <summary> <b>dolly-machine-translated</b> </summary>
270
+
271
+ ```json
272
+ {
273
+ "id": 2,
274
+ "inputs": "How to escape from a helicopter trapped in water ?",
275
+ "targets": "If you are ever trapped inside a helicopter while submerged in water, it’s best to try and remain calm until the cabin is completely underwater. It’s better to wait for pressure to be equalized, before you try to open the door or break the glass to escape.",
276
+ "language": "eng",
277
+ "script": "Latn",
278
+ "source_id": 6060,
279
+ }
280
+ ```
281
+ </details>
282
+
283
+ <details>
284
+ <summary> <b>dolly-human-edited</b> </summary>
285
+
286
+ ```json
287
+ {
288
+ "id": 2,
289
+ "inputs": "Comment peut-on s'échapper d'un hélicoptère piégé dans l'eau ?",
290
+ "targets": "-",
291
+ "language": "fra",
292
+ "script": "Latn",
293
+ "source_id": 6060,
294
+ }
295
+ ```
296
+ </details>
297
 
298
  ## Statistics
299