zhangguoxiao / bert-for-srl Goto Github PK

Where can I find the opensource ontonotes 5 dataset

你好，十分感谢你的开源，请问该项目使用的tensorflow版本号是多少？望解答，谢谢～

If one sentence have more than one predicates, they have same inputs.
Is this a problem?

Hi,
I was unable to reproduce the exact scores you had received on the English OntoNotes dataset. Are there certain hyperparameters not mentioned or what am I missing? Thanks.

Hello, thanks for your open source, after learning about your project, line 540 in bert_lstm_crf_srl.py still says "ValueError: None values not supported."Hope to get your answer soon.Best wishes!

感激你的开源分享，请问README.md中提到的“split learning strategy” 指的是什么呀，再次表示感激

Hello, Thank you for open sourcing the code! Perhaps I am doing wrong but I get an error when using a (large) dataset I created myself. In the README it says that the data should be formatted like: each line with two parts, one is BIO tags, one is the raw sentence with an annotated predicate, the two parts are splitted by "\t".

I figured that the problem might be how I provide a sentence with two predicates. Since tsv files are used as input to train the model I was wondering whether I formatted them correctly. It would be great if you gave an example of how the format should look like.

请问模型是怎么处理同一个输入，不同label的情况的，样例如下：
O O O O B_ARGM-ADJ V B_ARG1 I_ARG1 O O O O O O O O O O
In recent years , advanced education for professionals has become a hot topic in the business community .

B_ARGM-TMP I_ARGM-TMP I_ARGM-TMP O B_ARG1 I_ARG1 I_ARG1 I_ARG1 O V B_ARG2 I_ARG2 I_ARG2 I_ARG2 I_ARG2 I_ARG2 I_ARG2 O
In recent years , advanced education for professionals has become a hot topic in the business community .

使用您的代码，对于这种情况我得到的是相同的结果（类似于NER），请问您在实现中是怎么关注不同“verb”的