AI4Bharat Rasa (Marathi)

MH Specific

Expressive multilingual TTS dataset with neutral and emotional speech (6 Ekman emotions) for 22 Indian languages

Build a Marathi speech emotion recognition system for call center analytics to detect customer sentiment from voice.

Quick Start

from datasets import load_dataset
ds = load_dataset('ai4bharat/rasa', 'mr', split='train', streaming=True)
for i, ex in enumerate(ds):
    print(f"Emotion: {ex['emotion']}, Text: {ex['text'][:60]}...")
    if i >= 4: break

Modality

Speech + Text (Expressive TTS)

Size

~20 hrs/speaker; 400 hrs total (22 langs)

License

CC-BY 4.0

Format

WAV

Language

Update Frequency

static

Organization

AI4Bharat, IIT Madras

Schema

Field	Type	Description
audio	audio	Marathi emotional speech recording
emotion	string	Emotion label (happy, sad, angry, etc.)
text	string	Transcription of the spoken content

Build With This

Create a Marathi mental health screening tool that detects emotional distress signals from speech patterns

Develop an emotion-aware Marathi TTS system that generates speech with appropriate emotional intonation

Build a real-time call center dashboard that tracks customer emotion trends during Marathi-language support calls

AI Use Cases

Expressive TTSEmotion RecognitionVoice Acting

Related Datasets

AI4Bharat BhasaAnuvaad (Marathi)

Speech + Text (Translation)

AI4Bharat IndicVoices

speech+text

AI4Bharat IndicVoices-R

Speech + Text (TTS-ready)

AI4Bharat Kathbath

Speech + Text

Last verified: 2026-03-07