Combined Hindi ASR Dataset (athenasaurav/hindi_combined_all)

---

annotations_creators: [machine-generated] language_creators: [machine-generated] language: [hi] license: cc-by-nc-4.0 multilinguality: monolingual size_categories: [10K<n<100K] source_datasets: [] task_categories: [question-answering] task_ids: [open-domain-qa] pretty_name: Hindi QA Dataset

This repository contains a combined dataset for Hindi Automatic Speech Recognition (ASR), created by merging several publicly available Hugging Face datasets. The goal is to provide a larger, more diverse training corpus for Hindi ASR models. All audio data in this combined dataset has been resampled to 24000Hz.

Dataset Structure

The final dataset, pushed to athenasaurav/hindi_combined_all on the Hugging Face Hub, contains a single split: train.

Each example in the dataset has the following two columns, in this specific order:

1. audio: The audio data, which has been resampled to 24000Hz and is compatible with the Hugging Face datasets.Audio(sampling_rate=24000) feature.
2. text: The corresponding Hindi transcription for the audio, as a string.

All other columns from the original datasets have been dropped.

Source Datasets

The following Hugging Face datasets were combined to create this collection. The original specified columns for audio and text/transcription have been standardized to audio and text respectively, and all audio has been resampled to 24000Hz:

• cdactvm/hindi_dataset_v2 (original columns: audio, sentence)
• Sajjo/hindi_dataset (original columns: audio, sentence)
• RutwikShete/hindi_dataset (original columns: audio, transcription)
• shiva09/Hindi_New_Dataset (original columns: audio, sentence)
• TheAIchemist13/hindi_asr_dataset_2 (original columns: audio, transcription)
• damerajee/Hindi-audio-speech (original columns: audio, raw_text)
• MatrixSpeechAI/Common_voice_hindi_denoised (original columns: audio, transcription)
• SPRINGLab/IndicVoices-R_Hindi (original columns: audio, text)
• En1gma02/processed_hindi_speech_male_5hr (original columns: audio, text_hindi)
• MatrixSpeechAI/All_Hindi_ASR_v1.1 (original columns: audio, transcription)
• skywalker290/Hindi_TTS_M (original columns: audio, raw_text)

Usage

You can load this dataset using the HuggingFace Datasets library:

from datasets import load_dataset

ds = load_dataset("athenasaurav/hindi_combined_all", split="train")
print(ds[0])

License

This dataset is licensed under CC BY-NC 4.0. Non-commercial use only.