Datasets:

miracl
/

miracl

Tasks:

Sub-tasks:

document-retrieval

Languages:

Multilinguality:

Annotations Creators:

expert-generated

ArXiv:

License:

Dataset card Files Files and versions Community

crystina-z commited on Nov 2, 2022

Commit

c51474b

•

1 Parent(s): adff01c

testA

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -76,7 +76,7 @@ lang='ar'  # or any of the 16 languages
 miracl = datasets.load_dataset('miracl/miracl', lang, use_auth_token=True)
 # training set:
-for data in miracl['train']:  # or 'dev'
   query_id = data['query_id']
   query = data['query']
   positive_passages = data['positive_passages']
@@ -87,7 +87,7 @@ for data in miracl['train']:  # or 'dev'
     title = entry['title']
     text = entry['text']
 ```
-The structure is the same for `train` and `dev` set.
 Note that `negative_passages` are annotated by native speakers as well, instead of the non-positive passages from top-`k` retrieval results.

 miracl = datasets.load_dataset('miracl/miracl', lang, use_auth_token=True)
 # training set:
+for data in miracl['train']:  # or 'dev', 'testA'
   query_id = data['query_id']
   query = data['query']
   positive_passages = data['positive_passages']
     title = entry['title']
     text = entry['text']
 ```
+The structure is the same for `train`, `dev`, and `testA` set, where `testA` only exists for languages in Mr. TyDi (i.e., Arabic, Bengali, English, Finnish, Indonesian, Japanese, Korean, Russian, Swahili, Telugu, Thai).
 Note that `negative_passages` are annotated by native speakers as well, instead of the non-positive passages from top-`k` retrieval results.