Ruidan He

Hi, I did not include the coherence score computation in the code, you can find how it could be computed in Section 5.1.1 in the paper. However I do not...

You may read the paper section 5.2 and the script evaluation.py to understand how the F-scores are computed on test set.

The Keras version for this code is 1.2.1. If you want to keep the Keras version unchanged, you need to check which version of tensorflow will work with it. You...

@harpaj Thank you! That is very helpful :)

Thank you for the changes, but I think it would be better to change the numpy version and the syntax to be compatible with both python2 and python3.

Hi, I am not sure why you got the NaN loss, since I didn't encounter this issue with the current code or the code before changing batches_per_epoch to 1000. So...

Hi, I guess this is because that the aspect.log were generated with different random seed. Therefore, if you run the code on your machine, the words and clusters will be...

我当时实验使用的是keras 1.2.1， theano 0.9.0， README里还列出了其他一些dependencies的版本。你也可以参考其他的关于dependency的讨论 #5, #18.

words not in the vocab will be mapped to a special token "\".

you can check the format of the provided data and follow that.