dianavdavidson/MUCS-Hinglish

MUCS

Dataset Description

This dataset is a HuggingFace/Transformers compatible version of the MUCS 2021 Hinglish dataset. This dataset is part of the MUltilingual and Code-Switching ASR Challenges for Low Resource Indian Languages challenge, subtask 2.

As this dataset is in Hinglish, it contains codeswitching between Hindi and English. The original dataset was found here. In addition to making the dataset compatible for Transformers, preprocessing has been applied to indicate the ratio of Hindi words (ratio_hindi_words) and the ratio of English words (ratio_english_words) and the counts of words in each language in each audio segment. ratio_english_words_range indicate a 5% range for the raio of english words in a given audio segment. The Hinglish train and test datasets contain 89.86 hours and 5.18 hours of audio respectively (see here for more information about the dataset). All audio files are sampled at a rate of 16kHz. Vocabulary size for the Hinglish dataset is 17877. The Hinglish data are drawn from a repository of technical lectures on a diverse range of topics in computer science.

How to use & Supported Tasks

How to use

The datasets library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the load_dataset function.

For example, to download the Hindi config, simply specify the corresponding language config name (i.e., "hi" for Hindi):

from datasets import load_dataset
mucs = load_dataset("dianavdavidson/MUCS-Hinglish", split="train")

Using the datasets library, you can also stream the dataset on-the-fly by adding a streaming=True argument to the load_dataset function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.

from datasets import load_dataset
mucs = load_dataset("dianavdavidson/MUCS-Hinglish", split="train", streaming=True)
print(next(iter(mucs)))

audio_input = mucs["train"][0]["audio"]  # first decoded audio sample
transcription = mucs["train"][0]["transcript"]  # first transcription
# use `audio_input` and `transcription` to fine-tune your model for ASR