akshitab commited on
Commit
3dcfd2c
1 Parent(s): 67303e7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -12
README.md CHANGED
@@ -35,13 +35,29 @@
35
 
36
  ### Dataset Summary
37
 
38
- Machine translation research has traditionally placed an outsized focus on a limited number of languages - mostly belonging to the Indo-European family. Progress for many languages, some with millions of speakers, has been held back by data scarcity issues. An inspiring recent trend has been the increased attention paid to low-resource languages. However, these modeling efforts have been hindered by the lack of high quality, standardized evaluation benchmarks.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39
 
40
- For the second edition of the Large-Scale MT shared task, we aim to bring together the community on the topic of machine translation for a set of 24 African languages. We do so by introducing a high quality benchmark, paired with a fair and rigorous evaluation procedure.
41
 
42
  ### Supported Tasks and Leaderboards
43
 
44
- [Needs More Information]
45
 
46
  ### Languages
47
 
@@ -78,6 +94,8 @@ Colonial linguae francae: English - eng, French - fra
78
 
79
  ## Dataset Structure
80
 
 
 
81
  ### Data Instances
82
 
83
  The dataset contains 248 language pairs.
@@ -354,33 +372,33 @@ Example:
354
  ```
355
  ### Data Splits
356
 
357
- [Needs More Information]
358
 
359
  ## Dataset Creation
360
 
361
  ### Curation Rationale
362
 
363
- [Needs More Information]
364
 
365
  ### Source Data
366
 
367
  #### Initial Data Collection and Normalization
368
 
369
- [Needs More Information]
370
 
371
  #### Who are the source language producers?
372
 
373
- [Needs More Information]
374
 
375
  ### Annotations
376
 
377
  #### Annotation process
378
 
379
- The metadata for creation can be found here: https://github.com/facebookresearch/LASER/tree/main/data/wmt22_african
380
 
381
  #### Who are the annotators?
382
 
383
- [Needs More Information]
384
 
385
  ### Personal and Sensitive Information
386
 
@@ -390,11 +408,11 @@ The metadata for creation can be found here: https://github.com/facebookresearch
390
 
391
  ### Social Impact of Dataset
392
 
393
- [Needs More Information]
394
 
395
  ### Discussion of Biases
396
 
397
- [Needs More Information]
398
 
399
  ### Other Known Limitations
400
 
@@ -412,4 +430,4 @@ We are releasing this dataset under the terms of [CC-BY-NC](https://github.com/f
412
 
413
  ### Citation Information
414
 
415
- [Needs More Information]
35
 
36
  ### Dataset Summary
37
 
38
+ This dataset was created based on [metadata](https://github.com/facebookresearch/LASER/tree/main/data/wmt22_african) for mined bitext released by Meta AI. It contains bitext for 248 pairs for the African languages that are part of the [2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages](https://www.statmt.org/wmt22/large-scale-multilingual-translation-task.html).
39
+
40
+ #### How to use the data
41
+
42
+ There are two ways to access the data:
43
+
44
+ * Via the Hugging Face Python datasets library
45
+
46
+ ```
47
+ from datasets import load_dataset
48
+ dataset = load_dataset("allenai/wmt22_african")
49
+ ```
50
+
51
+ * Clone the git repo
52
+ ```
53
+ git lfs install
54
+ git clone https://huggingface.co/datasets/allenai/wmt22_african
55
+ ```
56
 
 
57
 
58
  ### Supported Tasks and Leaderboards
59
 
60
+ This dataset is one of resources allowed under the Constrained Track for the [2022 WMT Shared Task on Large Scale Machine Translation Evaluation for African Languages](https://www.statmt.org/wmt22/large-scale-multilingual-translation-task.html).
61
 
62
  ### Languages
63
 
94
 
95
  ## Dataset Structure
96
 
97
+ The dataset contains gzipped tab delimited text files for each direction. Each text file contains lines with parallel sentences.
98
+
99
  ### Data Instances
100
 
101
  The dataset contains 248 language pairs.
372
  ```
373
  ### Data Splits
374
 
375
+ The data is not split into train, dev, and test.
376
 
377
  ## Dataset Creation
378
 
379
  ### Curation Rationale
380
 
381
+ Parallel sentences from monolingual data in Common Crawl and ParaCrawl were identified via [Language-Agnostic Sentence Representation (LASER)](https://github.com/facebookresearch/LASER) encoders.
382
 
383
  ### Source Data
384
 
385
  #### Initial Data Collection and Normalization
386
 
387
+ Monolingual data was obtained from Common Crawl and ParaCrawl.
388
 
389
  #### Who are the source language producers?
390
 
391
+ Contributors to web text in Common Crawl and ParaCrawl.
392
 
393
  ### Annotations
394
 
395
  #### Annotation process
396
 
397
+ The data was not human annotated. The metadata used to create the dataset can be found here: https://github.com/facebookresearch/LASER/tree/main/data/wmt22_african
398
 
399
  #### Who are the annotators?
400
 
401
+ The data was not human annotated. Parallel text from Common Crawl and Para Crawl monolingual data were identified automatically via [LASER](https://github.com/facebookresearch/LASER) encoders.
402
 
403
  ### Personal and Sensitive Information
404
 
408
 
409
  ### Social Impact of Dataset
410
 
411
+ This dataset provides data for training machine learning systems for many languages that have low resources available for NLP.
412
 
413
  ### Discussion of Biases
414
 
415
+ Biases in the data have not been studied.
416
 
417
  ### Other Known Limitations
418
 
430
 
431
  ### Citation Information
432
 
433
+ Forthcoming research paper that describes the approach used to create the metadata. Citation Information will be updated with the paper information when that is available.