About the long text input in Scholar-XL dataset

[DavidLeexxxx]

Thanks for this excellent work in this specific domain. When I ran the CNN-base method on the Scholar-XL dataset which is a dataset for NER tasks, I came across with the token limitation errors, as we all know that bert can only accept an input token length within 512. I recognize that this error occurred pobaby because some inputs' lengths are larger than 512. This "out of token length" problem has also been mentioned in https://www.aminer.cn/scholar-profiling. However I counldn't find an solution or hyper parameters settings such as "max-length" about this problem in the source code(train_CNN.py). So could you please update your code about this or give an explanation about how to treat this problem in order to get the baseline results mentioned in this repo.