Dirk Weissenborn

bug in score sent

1

you pad the incoming sequence (https://github.com/smilli/kneser-ney/blob/master/kneser_ney.py#L147), but then go and use the original tuple (not padded) for scoring

Preliminary entity topic model implementation + Migration to scala 2.10 + Factorie NER spotters (trainable + trained models)

8

the entity topic model implementation is very simple and preliminary, but fast and working at the moment. It relies on the statistical backend for training (spotter+stores). It needs a lot...

question ids of reasoning types

3

It would be really nice if the question ids that require the reasoning types specified in the paper could be made available in order to make more detailed analyses.

see #384 .

https://blog.thedataincubator.com/2016/06/testing-jupyter-notebooks/

We proper support for ELMo that makes it easy to load and use. Integration should be easy given the TF implementation of the authors [1]. [1] https://github.com/allenai/bilm-tf

Spacy 2.0 slow for many short documents

2

... therefore we downgrade to 1.9 for the time being. See [1] for more info. [1] https://github.com/explosion/spaCy/issues/1508

see #361 currently solution is to fix the TF version in setup.py

Our current dataset setup is a bit cumbersome in some cases. We need a CLI that: 1) shows the user which dataset are available for which tasks 2) allows for...