allanj / ner_with_dependency

Hi there 👋

ner_with_dependency's Issues

Do you have conll2003 datasets with denpendency?

label not found error

at line 270 in main.py :
conf.build_label_idx(trains)
you use only trains labels to build a dict which may cause label not found in dict.It seem to be normal that training corpus should contain all kinds of labels,but it is not guaranteed that every corpus follow this limitation.The Interesting point is you use trains+devs+tests when you build word vocab and slot vocab. But you didn't do it in building label vocab.Is is a mistakes? Or It is based on another concern that i don't know?

A question about reproduction

Hello.

About the Chinese corpus, In table 4, you reported the naive BiLSTM-CRF (L = 2) can reach 76.61 (F1 score). I have tried to reproduce that by using my own implementation, but I can not get that number.
Could you please tell me how to reproduce that with your implementation? What's the command to do that?

About Chinese corpus

How to transfer a Chinese sentence to conllx format?

Potential bugs when running "dggcn" mode

Hi, I attempted to run your codes in the "dggcn" mode, but I got a weird error "RuntimeError: CUDA error: device-side assert triggered" in this line:

ner_with_dependency/model/deplabel_gcn.py

Line 76 in a1117dd

dep_embs = dep_embs.squeeze(3) * adj_matrix

and this line:

ner_with_dependency/model/deplabel_gcn.py

Line 79 in a1117dd

dep_denom = dep_embs.sum(2).unsqueeze(2) + self_val

help ，i don't know how to use your code

I am a beginner and I think your code is very helpful to me. I want to know how I can successfully run your code. I think I am missing some files, but I don’t know exactly which ones and where to get them

Convert the constituency trees into the Stanford dependency trees

The paper says "We convert the constituency trees into the Stanford dependency trees using the rulebased tool by Stanford CoreNLP." Could you please share the code that you preprocess the data to adapt to the input format of Stanford CoreNLP tools? Thanks a lot in advance!

conllU

Could you please tell me how to transform OntoNotes 5 to CoNLL-U format? Thank you very much!!

How to use Chinese data

I saw that you use Chinese corpus on OntoNotes ,so i am wandering how to use Chinese corpus.When i use Chinese corpus , it seem that some words will be packed together ,which may have different slot.How do you cope with this problem. Your rapid reply will be highly appreciated. Thx.

Is the F1 value in your paper the best result?

Hi! I'm bothering you again ...
I recently read your paper and try to reproduce your results. The result for SemEval-2010 Task 1 in your paper is as follows:

and I get the highest F1 82.83 and 84.14 respectively different from your result 82.19(DGLSTM-CRF L=1) and 83.47 (DGLSTM-CRF L=2).

I wonder if my config is not set correctly or do you give conservative values？

I just run
python main.py --device cuda:0 --dep_model dglstm --momentum 0.9 --lr_decay 0.02 --dataset spanish --embedding_file data/cc.es.300.vec