model2me/scienceqa-multilingual-hindi
#ScienceQA Hindi Translation Dataset ##Dataset Description This dataset is a Hindi-translated version of the original ScienceQA dataset. It includes multiple-choice science questions, with fields for: Images (optional visual context), Hints (optional support text), English questions and their Hindi translations, Multiple answer choices, Correct answers. This translation is intended to support multilingual education research, question-answering in Hindi, and fairness studies in multilingual… See the full description on the dataset page: https://huggingface.co/datasets/model2me/scienceqa-multilingual-hindi.
#ScienceQA Hindi Translation Dataset
##Dataset Description
This dataset is a Hindi-translated version of the original ScienceQA dataset. It includes multiple-choice science questions, with fields for:
- Images (optional visual context),
- Hints (optional support text),
- English questions and their Hindi translations,
- Multiple answer choices,
- Correct answers.
This translation is intended to support multilingual education research, question-answering in Hindi, and fairness studies in multilingual NLP.
##Translation Process
- The English questions were translated to Hindi using the deep-translator package with the Google Translate API.
- Translations were run in parallel using joblib.Parallel to speed up the process and mitigate rate limits.
- If a translation failed (e.g., due to rate limiting), a retry mechanism was used. Unresolved failures were marked with [TRANSLATION FAILED].
Example logic: python try: translated = GoogleTranslator(source='en', target='hi').translate(text) except: translated = "[TRANSLATION FAILED]"