I read your paper and curious about results, but want to replicate it by using your case study data. What data did you use for your case study?

Dear authors,

Could you please share the data preprocess code with us? Because the data format is differenet with the original .xml file.

If you could share this preprocess file, we can use your trained model to do some weakly supervise work.

Thanks a lot.

Bug: CRF not defined

When I run the train.py script with the use_crf flag set to true (
python main.py --ds 14res --use_crf True) I get the following error: NameError: name 'CRF' is not defined

This error is thrown here:

Training with the --use_crf flag set to False works as expected without errors.
Am I missing a package or something?

您好，
根据
Dependency:
python3
pytorch 0.4
当前发现很难跑通代码。尤其是torchtext的版本不匹配。
请问有更详细的Dependency文件吗

I find a strang case with s_id=3155 (line 86).

3155	They are often crowded on the weekends but they are efficient and accurate with their service .	They\O are\O often\O crowded\O on\O the\O weekends\O but\O they\O are\O efficient\O and\O accurate\O with\O their\O service\B .\O	They\O are\O often\O crowded\O on\O the\O weekends\O but\O they\O are\O efficient\B and\O accurate\B with\O their\O service\O .\O

The opinion_target_tag and the opinion_word_tag are the same.
Is it normal?

同学你好，这几天拜读了您的这篇Target-oriented Opinion Words Extraction with Target-fused Neural Sequence Labeling，我有一些细节方面的疑问，比如：
假设在一句话中出现重复的target应该怎么办呢？比如下面这个例子：
To be honest, most food are delicious! And those desserts are my favourites. Besides, other green food is also popular among people.
这个例子里的target:food在文中出现了两个地方，那应该如何计算Inword和Outword LSTM呢？文中的例子好像默认一句话中target只在一处出现。

另一个疑问就是如果句中没有显式的target那应该如何解决呢，例如：
The cappuccino is quite bitter and lattes is even worse. I bet other coffee is bad as well.
如果target是drinks,但是文中并没有出现，那应该如何计算呢？

可能文中有些细节我没读到位导致了这些困惑，等待您的解答~

本文提出的TOWE任务，是给定opinion target ，找句子中的 opinion word，然后成对提取出来。那么在test时，是怎么训练呢，也需要在test集中为每个句子标注opinion target吗？

In your code there are multiple references to ELMo, indicating that you experimented with contextual embeddings.

Can you share any of your results using ELMo embeddings?
I am currently getting f1-scores of ~85 on 14res and 16res using Flair embeddings instead of GloVe.

I would be very interested in hearing about your experiments. Thank you!