Hugging Face's logo

Datasets:
ShimogaAIteam
/
conversational_kannada_stt

Tasks:
Automatic Speech Recognition
Modalities:
Text
Formats:
parquet
Languages:
Kannada
Size:
1K - 10K
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
1
Dataset Viewer (First 5GB)
Auto-converted to Parquet
sentence
stringlengths
4
335
input_features
listlengths
80
80
labels
listlengths
448
448
audio_array
listlengths
68.4k
240k
"ಹೆಸರು ನನಗೆ ಗೊತ್ತಿಲ್ಲ. ಆದರೆ ರಾಮನ ಹೆಸರು(...TRUNCATED)
[[0.5163508057594299,0.4196360111236572,0.43835675716400146,0.4979405403137207,0.44774329662323,0.42(...TRUNCATED)
[50258,50306,50359,50363,11891,117,24998,229,11891,111,11891,123,11891,99,44833,11891,110,24998,223,(...TRUNCATED)
[-0.021209716796875,-0.02789306640625,-0.028411865234375,-0.02874755859375,-0.023651123046875,-0.021(...TRUNCATED)
"ತಯಾರು ಮಾಡಿಕೊಡಿ. ಆದರೆ ಮಕ್ಕಳಿಗೆ ಹೊರಗಿನ (...TRUNCATED)
[[0.2660527229309082,0.11633789539337158,0.23341184854507446,0.369409441947937,0.3666398525238037,0.(...TRUNCATED)
[50258,50306,50359,50363,11891,106,11891,101,24998,229,11891,110,24998,228,34725,106,11891,243,44833(...TRUNCATED)
[0.003143310546875,0.0062255859375,0.005340576171875,0.00714111328125,0.006378173828125,0.0057678222(...TRUNCATED)
"ಒಂದು ದೊಡ್ಡ ಗಾತ್ರದ ಈರುಳ್ಳಿ, ಸಣ್ಣದಾಗಿ ಹ(...TRUNCATED)
[[0.6244785189628601,0.4676451086997986,0.5796871781349182,0.5347731709480286,0.5967820882797241,0.5(...TRUNCATED)
[50258,50306,50359,50363,11891,250,24998,222,11891,108,11891,123,11891,245,24998,228,34725,248,11891(...TRUNCATED)
[-0.04022216796875,-0.07183837890625,-0.06793212890625,-0.0755615234375,-0.078582763671875,-0.081909(...TRUNCATED)
"ಎಂಬುದು ನನಗೆ ಗೊತ್ತು. ಕೊರಳಿಗೆ ಒಂದು ಸರ, ಕ(...TRUNCATED)
[[0.3833656907081604,-0.04757857322692871,0.020716190338134766,-0.17618095874786377,0.20872896909713(...TRUNCATED)
[50258,50306,50359,50363,11891,228,11891,255,11891,108,11891,96,11891,245,11891,111,11891,101,44833,(...TRUNCATED)
[0.027618408203125,0.045166015625,0.0416259765625,0.039764404296875,0.040069580078125,0.044616699218(...TRUNCATED)
"ಕ್ಯಾಲ್ಸಿಯಂ ಇರುತ್ತದೆ. ಮತ್ತೆ ಈ ಸೂರ್ಯಕಾ(...TRUNCATED)
[[0.26622045040130615,0.06627088785171509,-0.12826502323150635,0.2995401620864868,0.3330222964286804(...TRUNCATED)
[50258,50306,50359,50363,11891,117,24998,228,11891,248,44833,11891,248,24998,223,34725,243,44833,118(...TRUNCATED)
[0.044281005859375,0.075531005859375,0.06689453125,0.061981201171875,0.04998779296875,0.038787841796(...TRUNCATED)
"ತನ್ನ ಅಳಿಯ ಪುಟ್ಟನ ಹೆಸರನ್ನ ಸೂಚಿಸುವಳು. ಮ(...TRUNCATED)
[[0.9288715124130249,0.5469410419464111,0.30973494052886963,0.5291587114334106,0.42080360651016235,0(...TRUNCATED)
[50258,50306,50359,50363,11891,97,11891,101,44833,11891,101,34725,106,11891,245,11891,111,11891,101,(...TRUNCATED)
[-0.027740478515625,-0.03033447265625,0.0189208984375,0.06024169921875,0.090576171875,0.113189697265(...TRUNCATED)
"ಮುಸಲ್ಮಾನರಿಂದ ಆದದ್ದಲ್ಲ, ಶೈವ ವೈಷ್ಣವರ ಜ(...TRUNCATED)
[[0.1609455943107605,0.3709794878959656,0.20189183950424194,0.17132633924484253,0.3939365744590759,0(...TRUNCATED)
[50258,50306,50359,50363,11891,227,11891,99,11891,108,34725,103,44833,11891,108,11891,243,11891,122,(...TRUNCATED)
[-0.0126953125,-0.019317626953125,-0.016571044921875,-0.016265869140625,-0.0142822265625,-0.01251220(...TRUNCATED)
"ಯಾರು ಏನೂ ಮಾತಾಡಿಲ್ಲ ಅಂದೆ. ನೋಡ್ರೋ ನೀವು (...TRUNCATED)
[[0.8119224905967712,0.2767765522003174,0.19549107551574707,0.4188305139541626,0.4664314389228821,0.(...TRUNCATED)
[50258,50306,50359,50363,11891,116,44833,11891,113,11891,122,11891,106,24998,222,11891,250,11891,123(...TRUNCATED)
[-0.121063232421875,-0.15338134765625,-0.0816650390625,-0.067474365234375,-0.047882080078125,-0.0560(...TRUNCATED)
"ಒಂದು ಬಿಟ್ಟು ಅಳೋಕೆ ಶುರು ಮಾಡಿದ್ಳು. ಏನು? (...TRUNCATED)
[[0.5999376773834229,0.2327912449836731,0.34719210863113403,0.12998723983764648,0.3469204902648926,0(...TRUNCATED)
[50258,50306,50359,50363,11891,227,11891,224,11891,97,34725,243,24998,229,11891,111,11891,123,11891,(...TRUNCATED)
[0.084259033203125,0.091552734375,0.0164794921875,-0.029815673828125,-0.0728759765625,-0.09320068359(...TRUNCATED)
"ಅಂತ ನೀವು ವೀಕ್ ಅಂತ ಅನ್ಕೋಬೇಡಿ. ನಿಮಗೆ ಏನ(...TRUNCATED)
[[0.798835813999176,0.1867750883102417,0.044251441955566406,-0.22279775142669678,-0.1585506200790405(...TRUNCATED)
[50258,50306,50359,50363,11891,227,11891,115,44833,11891,253,24998,223,34725,253,11891,104,44833,118(...TRUNCATED)
[0.11474609375,0.165283203125,0.1170654296875,0.097808837890625,0.076324462890625,0.06317138671875,0(...TRUNCATED)
End of preview. Expand in Data Studio

Conversational Kannada STT

This dataset contains corrected transcriptions of conversational Kannada speech,
prepared specifically for fine-tuning Whisper models on conversational and dialectal Kannada.

Unlike many ASR datasets, this release provides pre-computed Whisper input features (log-Mel spectrograms)
so you can train/fine-tune Whisper models without raw audio processing.

Dataset Creation

  • Source Audio: Publicly available YouTube videos in Kannada.
  • Initial Transcripts: Extracted using YouTube’s auto-transcription system.
  • Corrections:
    • First-pass corrections with Gemini Pro for common errors.
    • Final manual corrections by ShimogaAIteam, focusing on dialects & colloquial usage.

This process ensures the dataset is closer to real-world conversational Kannada.

Dataset Details

  • Language: Kannada (kn)
  • Domain: Conversational & Dialectal Speech
  • Rows: ~5K
  • Audio: ❌ Not included
  • Features: ✅ Pre-computed Whisper log-Mel spectrograms

Columns

Column Description
sentence Corrected transcription in Kannada
input_features Pre-computed Whisper log-Mel spectrogram features
labels Tokenized labels for Whisper training

Intended Use

  • Fine-tuning Whisper for Kannada conversational ASR
  • Coverage of dialects and colloquial Kannada
  • Research on low-resource ASR in Indian languages

Usage Example 

from datasets import load_dataset

dataset = load_dataset("ShimogaAIteam/conversational_kannada_stt")

print(dataset)
print(dataset["train"][0]["sentence"])
Citation
If you use this dataset, please cite:

bibtex
Copy code
@dataset{shimogaai_conversational_kannada_stt,
  title        = {Conversational Kannada STT (Corrected)},
  author       = {ShimogaAIteam},
  year         = {2025},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/ShimogaAIteam/conversational_kannada_stt}
}
Acknowledgements
ShimogaAIteam for manual corrections & dataset preparation

YouTube for providing source audio & transcripts

Gemini Pro for transcript refinement assistance

The open-source Whisper community for enabling Kannada ASR research
Downloads last month
71
Size of the auto-converted Parquet files (First 5GB):
1.5 GB
Number of rows (First 5GB):
1,336
Estimated number of rows:
9,138