metadata
license: mit
This dataset by Mozilla Common Voice (https://commonvoice.mozilla.org/en/datasets) is crafted by Udyan Sachdev
Voice datasets play a pivotal role in training and evaluating speech-to-text models, influencing advancements in natural language processing. This dataset outlines the creation of a comprehensive text dataset from 40,571 MP3 audio files sourced from the Common Voice project. The dataset aims to serve as a benchmark for training and evaluating speech-to-text models in English, French, and Spanish languages, leveraging the OpenAI Whisper-large-v3 model.
Data Details: Common Voice Delta Segment: • Size: 1.28 GB (40,571 MP3 audio files) • Duration: 68 recorded hours, 48 validated hours • Voices: 750 unique voices • Format: MP3 audio