Chinese Hypotheses Paradise dataset (ChineseHP)

This dataset is published with the paper Pinyin Regularization in Error Correction for Chinese Speech Recognition with Large Language Models in Interspeech 2024.

• The N-best for both text and Pinyin in data/ are separated by '|'. Pinyin is directly transcribed from text.

• papar_train_list contains ids of subset of the training set used in the paper.