Update README.md
Browse files
README.md
CHANGED
|
@@ -18,6 +18,44 @@ IndicGenBench is a multilingual, multi-way parallel benchmark for measuring lang
|
|
| 18 |
|
| 19 |
We extend existing datasets in Cross-lingual Summarization (CrossSum), Machine Translation (FLORES), Multi-lingual Question Answering (XQuAD), and Cross-lingual Question Answering (XorQA) by collecting human translations for English examples into the target Indic languages. Please see the paper for details.
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
### Dataset Description
|
| 22 |
|
| 23 |
<!-- Provide a longer summary of what this dataset is. -->
|
|
|
|
| 18 |
|
| 19 |
We extend existing datasets in Cross-lingual Summarization (CrossSum), Machine Translation (FLORES), Multi-lingual Question Answering (XQuAD), and Cross-lingual Question Answering (XorQA) by collecting human translations for English examples into the target Indic languages. Please see the paper for details.
|
| 20 |
|
| 21 |
+
### Dataset Structure
|
| 22 |
+
|
| 23 |
+
#### Dataset Fields
|
| 24 |
+
|
| 25 |
+
- text: The paragraph which is to be summarized. This paragraph is in English.
|
| 26 |
+
- summary: summary of the paragraph in the langauge specified by the "lang" field.
|
| 27 |
+
- lang: language code of the targeted language. 'summary' is in this language.
|
| 28 |
+
|
| 29 |
+
#### Dataset Instances
|
| 30 |
+
|
| 31 |
+
```
|
| 32 |
+
{
|
| 33 |
+
'lang': 'as',
|
| 34 |
+
'source_url': 'https://www.bbc.com/news/technology-39193008',
|
| 35 |
+
'summary': 'Wikileaksএ এনে কিছু সবিশেষ প্ৰকাশ কৰিছে, যিবোৰ ইয়াৰ মতে এইবোৰ '
|
| 36 |
+
'হৈছে চিআইএৰ দ্বাৰা ব্যৱহৃত বিস্তৃত পৰিসৰৰ হেকিং সঁজুলি।',
|
| 37 |
+
'target_url': 'https://www.bbc.com/news/technology-39193008',
|
| 38 |
+
'text': 'By Leo KelionTechnology desk editor The alleged cyber-weapons are '
|
| 39 |
+
'said to include malware that targets Windows, Android, iOS, OSX and '
|
| 40 |
+
'Linux computers as well as internet routers. Some of the software is '
|
| 41 |
+
"Reported [...TRUNCATED…].'
|
| 42 |
+
}
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
#### Dataset Usage Instructions
|
| 46 |
+
|
| 47 |
+
for loading the dataset and printing the first 5 examples from the validation set:
|
| 48 |
+
|
| 49 |
+
```
|
| 50 |
+
dataset = load_dataset('google/IndicGenBench_crosssum_in', field="examples")
|
| 51 |
+
|
| 52 |
+
for i, batch in enumerate(dataset['validation']):
|
| 53 |
+
pprint(batch)
|
| 54 |
+
if i==4:
|
| 55 |
+
break
|
| 56 |
+
```
|
| 57 |
+
|
| 58 |
+
|
| 59 |
### Dataset Description
|
| 60 |
|
| 61 |
<!-- Provide a longer summary of what this dataset is. -->
|