thecharm/boundary-aware-nested-ner

Code for EMNLP 2019 paper "A Boundary-aware Neural Model for Nested Named Entity Recognition"

boundary-aware-nested-ner's Issues

getting error while training/evaluating on my custom dataset

getting following error:

all 0 tags, no evaluating this epoch
all 0 tags, no evaluating this epoch
all 0 tags, no evaluating this epoch

ValueError: Number of classes, 0, does not match size of target_names, 2. Try specifying the labels parameter

The file is missing

留名

Need some help for the embedding.npy

Thank you for the excellent work!

I am tring to run the train.py. But the EMBED_URL file in the train, which is needed to bulid the model is missing. This file is not in the google driver bio_nlp_vec.tar.gz. Would you provide this file?

In addition, where can I get the iob2 format 'Germ' dataset? The link in README is bad.

Thank you again.

region_labels

boundary-aware-nested-ner/Our_boundary-aware_model/dataset.py

Line 57 in 477bd13

for start in range(0, length):

boundary-aware-nested-ner/Our_boundary-aware_model/dataset.py

Line 62 in 477bd13

region_labels.append(self.label_ids[records[(start, end+1)]] if (start, end+1) in records else 0)

这两行之间实现的region_labels代表什么意思啊？谢谢，不太明白嵌套试实体的标签怎么表示的，希望赐教。

"IndexError: Target 12 is out of bounds."

wikipedia.bin在windows系统中咋用的?

大佬你好，我是科研小白，感谢您分享论文中的源码，在neural_layered_model中，我还是没看懂那个wikipedia200.bin咋用的。您给的配置环境是Ubuntu中，我是用的windows，所以在这个模型中报错不断，希望大佬看见消息，可以指点一下，非常感谢，谢谢你分享源码，又给予指导，非常感谢。

RuntimeError: There were no tensor arguments to this function

model CatRepr 结构在处理迭代数据的过程（也就是说报错之前的 batch size 数据不存在这个问题）突然有某个 batch 的数据爆出这个这个错误，传入 CatRepr 的 Tensor 数据变成了空列表。
还有谁遇到了这个问题嘛? 我检查了数据结构，但是我死活不能发现问题出在了哪里。 帮帮我大佬们~

how to extract nested entities？

when train or evaluate，the sentence_labels has no nested label，is flat，so I want to kown，how to extract nested entities when predict？

当训练或者验证的时候， sentence_labels这个变量遇到标注为1和2的就跳过了，是扁平的不是嵌套的，那么模型训练完后，如何提取嵌套实体呢？

GermEval bad tokens

Hi, reviewing the GermEval file, I noticed that there are several tokens that are empty strings, associated to the O O tags. Did you have no problem processing these tokens?

What‘s the "vecob.json" file and "char_vecob.json" type

There was a slight error while I was running the code,and I want to know the format of embedding document,Let me replace it with a new file.Because the embedding document provided by you requires VPN to download it.