luanyi / dygie Goto Github PK

I see preprocessing scripts for ACE and ontonotes, but where are the scripts for the wet lab dataset?

I was reading the code and got confused what ``head_word_emb'' means, is it the embeddings of entity head? Is it used in NER? Thanks!

The current run.zsh has issues for preprocessing in ace2004.

To fix it,

1. change java -cp ".:../stanford-corenlp-full-2015-04-20/* to java -cp ".:../common/stanford-corenlp-full-2015-04-20/*
2. delete &>! log & in adjust offsets

Hi,

Does this line replace 'test' by 'train' wrongly?

Best,
Deming

hello,I am trying to review at the srl_modle.py code and found this. In code line 145: mention_emb = attended_span_emb. does this code deactivates line 148 mention_emb = f * attended_span_emb + (1 - f) * mention_emb # [k, emb] ### ### ###

[_**.**_](https://github.com/luanyi/DyGIE/blob/master/model1/srl_model.py

Can this model extract inter-sentence relations b/w two entities? If so, how should one prepare data for the same?

Hi @luanyi ,
Thank you for sharing this project. The paper is very interesting and the results are impressive.

I am trying to follow the README just to reproduce the results from the paper, but there are many errors and mistakes.
I am fixing some of them, but others always pop up.
Most issues are related to incorrect data paths.

I see that others experience similar issues and are usually not getting any responses.
So - is this project active and maintained? Can I ask more questions about running, training, etc?

Thanks!
Uri

Thank you for this program, it's really helpful.

I have upgrade the nvidia driver to cuda-9.2 long ago which is not supported by tensorflow 1.8. Is it possible to run these programs with newer tensorflow? I guess tensorflow 2.0 and python3 will be better than current version.

And there is a missing file 'scripts/get_char_vocab.py'. Not sure what it does though.

I preprocess the ACE2005 corpus through your code but found some issues. The first issue is that in ace2json.py line93 the print function is in python2. I manually change the print function and run ace2json.py file. But it shows that I miss some files. I found that in ace2005/text/CNN_IP_20030408.1600.04.txt.conll the Standford annotator would wrongly add one more period in line 87~88 the word "SR..". there are two periods. Can you check this issue? Thanks for your contribution.

Hello,

In your paper you only specify the criterion to consider an entity as correct, and not a relation.
As I understand by a quick look at your code in model1/relation_metrics.py you consider a relation as correct if the relation type is correct along with the spans of its two arguments.
That is without considering the predicted entity type of the arguments.

If so, you use what (Bekoulis 2018) refers to as the "Boundaries" evaluation setting.
You cannot directly compare to previous works that take into account the entity type in the "Strict" evaluation setting as defined by (Bekoulis 2018).
As far as I know, (Li and Ji 2014) is the only related work using this "Boundaries" evaluation setting.

FYI (Sanh 2019) also uses a different metric and its scores are already not comparable to previous work, as pointed out in this issue.

(Bekoulis 2018) = "Joint entity recognition and relation extraction as a multi-head selection problem"

Best regards,

hello, I found these best models cannot be downloaded,can you share new address?