# Attention-Based Speech Recognition — Data

来源：https://www.kaggle.com/competitions/11-785-f22-hw4p2/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

You must use the train data to train the model. Data in the dev folder are to be used only for
validation and not for training. You must transcribe the utterances in the test folder and write them out into submission.csv
and upload your results to Kaggle.