Megathon 2022 code dump
Accent Detection: Given an audio clip of a person speaking in English, you are required to identify the native language of the speaker based on their accent when they are speaking in English.
You have been given a dataset of audio files of various people - of different native languages - speaking something. You can use these to train your model. The dataset is organized such that all audio clips of a speaker of a particular native language are placed within a directory with the name of the native language of the speaker. For example: All audio clips placed inside the “afrikaans” directory contain audio clips of people whose native language is afrikaans.
Audio data was taken from Speech Accent Archive https://www.kaggle.com/datasets/rtatman/speech-accent-archive
You will be evaluated on the basis of:
- The accuracy scores (weighted F-1 score) of your model
- Novelty (in any form like model architecture, training methodology or ideation)
- Clarity of presentation