Hugging Face's logo

Datasets:
alibashir
/
CommonsenseQA

Tasks:
Question Answering
Modalities:
Text
Formats:
json
Languages:
English
German
French
Size:
10K - 100K
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
Community
1
alibashir commited on
Commit
964699b
·
verified ·
1 Parent(s): 12506e7

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +20 -6
README.md CHANGED
@@ -44,6 +44,10 @@ configs:
44
  data_files:
45
  - split: train
46
  path: data/multi-lang/train.jsonl
 
 
 
 
47
  ---
48
 
49
  # CommonsenseQA Multilingual
@@ -55,12 +59,19 @@ Multilingual version of CommonsenseQA with GPT-4o-mini/GPT-5-mini generated Chai
55
  - German (de) - translated
56
  - French (fr) - translated
57
  - Chinese (zh) - translated
58
- - **Urdu (ur) - translated (NEW)**
59
 
60
  ## Splits
 
 
61
  - **train**: 8,196 examples per language
62
  - **validation**: 1,221 examples per language
63
- - **multi-lang**: 8,196 examples (mixed, ~2,049 per language for 4 languages, non-overlapping questions)
 
 
 
 
 
64
 
65
  ## Usage
66
 
@@ -68,19 +79,22 @@ Multilingual version of CommonsenseQA with GPT-4o-mini/GPT-5-mini generated Chai
68
  from datasets import load_dataset
69
 
70
  # Load a specific language
71
- ds = load_dataset("alibashir/CommonsenseQA", "ur") # Urdu
72
 
73
- # Load multi-lang (mixed languages)
74
  ds = load_dataset("alibashir/CommonsenseQA", "multi-lang")
 
 
 
75
  ```
76
 
77
  ## Fields
78
  - `question`: The question with answer choices
79
  - `cot`: Chain-of-thought explanation
80
  - `answer`: Correct answer (A-E)
81
- - `lang`: Language code (only in multi-lang split)
82
 
83
  ## Changelog
84
- - **v3**: Added Urdu (ur) translations
85
  - **v2**: Removed extra conclusion sentences from translations
86
  - **v1**: Initial release with en, de, fr, zh
 
44
  data_files:
45
  - split: train
46
  path: data/multi-lang/train.jsonl
47
+ - config_name: multi-lang-with-urdu
48
+ data_files:
49
+ - split: train
50
+ path: data/multi-lang-with-urdu/train.jsonl
51
  ---
52
 
53
  # CommonsenseQA Multilingual
 
59
  - German (de) - translated
60
  - French (fr) - translated
61
  - Chinese (zh) - translated
62
+ - Urdu (ur) - translated
63
 
64
  ## Splits
65
+
66
+ ### Single Language Splits
67
  - **train**: 8,196 examples per language
68
  - **validation**: 1,221 examples per language
69
+
70
+ ### Multi-Language Splits
71
+ - **multi-lang**: 8,196 examples (4 languages: en, fr, de, zh - ~2,049 per language, non-overlapping)
72
+ - **multi-lang-with-urdu**: 8,195 examples (5 languages: en, fr, de, zh, ur - 1,639 per language, non-overlapping)
73
+
74
+ **Important**: In multi-language splits, each question appears in ONLY ONE language. There is no overlap.
75
 
76
  ## Usage
77
 
 
79
  from datasets import load_dataset
80
 
81
  # Load a specific language
82
+ ds = load_dataset("alibashir/CommonsenseQA", "ur")
83
 
84
+ # Load multi-lang (4 languages)
85
  ds = load_dataset("alibashir/CommonsenseQA", "multi-lang")
86
+
87
+ # Load multi-lang-with-urdu (5 languages)
88
+ ds = load_dataset("alibashir/CommonsenseQA", "multi-lang-with-urdu")
89
  ```
90
 
91
  ## Fields
92
  - `question`: The question with answer choices
93
  - `cot`: Chain-of-thought explanation
94
  - `answer`: Correct answer (A-E)
95
+ - `lang`: Language code (only in multi-lang splits)
96
 
97
  ## Changelog
98
+ - **v3**: Added Urdu (ur) translations and multi-lang-with-urdu split
99
  - **v2**: Removed extra conclusion sentences from translations
100
  - **v1**: Initial release with en, de, fr, zh