some thing about the environment configuration

[bbruceyuan]

hello

I am really interest in your great work in AM. Some problems confused me a long time when I tried my best to reproduce the result of your paper.

• Question one: When I run the model BiLSTM-CNN-CRF, I found that the repo BiLSTM-CNN-CRF: https://github.com/XuezheMax/LasagneNLP was not supported any more, so I use that author's new repo NeuroNLP2. Then there was problem like this. maybe it's because label is too much. do you have any idea?
• Question two: as I found that I can't use the new repo "NeroNLP2", I extract your /BiLSTM-CNN-CRF/acl2017.tgz, it's not ok for theano's new edition . So can u share your requirements of the project with specific edition of the used package? especially the theano's edition because Theano's api had been changed.

thank you very much

[SteffenEger]

Hi,

thanks for your interest.

Concerning problem 1: If they set up something new, I don't unfortunately know what to do in case it doesn't work anymore.

Concerning problem 2: Does it run with theano version 0.9.0? I'm unfortunately not sure anymore which version we used at the time.

Finally, I recommend you try this implementation in the mean time: https://github.com/UKPLab/emnlp2017-bilstm-cnn-crf

I think it's both faster and more user-friendly than the other implementations, but implementing the same model.