# ভাষা-বিচিত্রা: ASR for Regional Dialects — Data

来源：https://www.kaggle.com/competitions/ben10/data

采集时间：2026-10-18T14:12:32Z

The recordings are 16kHz *.wav audio files. The data are coming from 373 different speakers from 10 regions in Bangladesh (rangpur, kishoreganj, narail, chittagong, narsingdi, tangail, habiganj, barishal, sylhet, sandwip) There are in total 13610 samples in the train set (63:08:42 hrs) and 1703 samples in the phase 1 test set (8:01:11 hrs).

<> notation is present throughout the train and test sets where the there exists a word, but not intelligible (due to noise, slurred speech, overlap etc). Newline operators (\n) are present only in the training data but not in the test ground truth files. 

The unicode of the validation (Phase 1 test set)/test (Phase 2 test set) dataset ground truth is normalized using [bnunicodenormalizer](https://github.com/mnansary/bnUnicodeNormalizer). 

## Files

* The **16_kHz_train_audio** folder contains the train files, and the **16_kHz_valid_audio** folder contains the phase 1 test set files.  
*   **train.csv** - the training set transcriptions
*   **sample_submission.csv** - a sample submission file in the correct format.

## Columns

*   `id` - wav file name. (lorem1.wav)
*   `sentence` - transcript of file.