# HW5: Unsupervised Speech Recognition — Data

来源：https://www.kaggle.com/competitions/11785-fall2021-hw5/data

采集时间：2026-10-19T06:31:07Z

- phoneme_map.json: maps each phoneme to a unique character. You need to use it to convert your phoneme sequence into character strings for comparison. Ignore any phonemes that's not present in this json file
- test.lengths and test.npy: test split data
- sample_submission.csv: sample submission file

The rest of the files are the training and validation set that is the same as the ones released on Google Drive.