liuwangwang

比如说 今天，父母等

 for replace , when we calculate attention score of position i , we don't consider the token w(i). at the first layer , I think it is no problem,...

I can't understand the result in README.md file, "It turn out that Tucker decomposition yields lower accuracy loss than CP decomposition in my experiments, so the results below are all from Tucker decomposition."

1

"It turn out that Tucker decomposition yields lower accuracy loss than CP decomposition in my experiments, so the results below are all from Tucker decomposition." what means "Tucker decomposition yields...

can you share a dataset that can make your code run correctly？

感谢开源！ 我知道，要使用data_process文件夹下面的代码？ 但是对于该文件夹中代码的使用顺序不太清楚。另外，参数phonetics_vocab_dir，从哪里获取呢？ 期待你的回复！谢谢！

can you release your LM models ？

1

sh pipeline.sh [-t test-sentences.txt] \ [-n native_english.kenlm] \ [-l learner_english.kenlm] \ [-w working/directory] how to get the native_english.kenlm file and learner_english.kenlm file ? can you release these file for testing