deep_qa
deep_qa copied to clipboard
A deep NLP library, based on Keras / tf, focused on question answering (but useful for other NLP too)
Firstly, Much thanks to this great project, which is what I would like to do; I'll continuously watch, use, and even contribute to this project. But when I want to...
I'm pretty sure this won't work on Windows at the moment, because we have plenty of places where `/` is hard-coded, instead of using `os.path.join`. This should be fixed. It's...
Looks like our build scripts for CircleCI [intentionally clear the cache](https://github.com/allenai/deep_qa/blob/7f008e47696cdd526ddb88c17f850e916cd46487/build_tools/circle/install_doc_requirements.sh#L7-L8) before building the docs, which makes things take longer than they probably need to. It didn't use to matter...
This should be doable with Keras, as shown here: https://medium.com/@kuza55/transparent-multi-gpu-training-on-tensorflow-with-keras-8b0016fd9012. Matt Peters also got this working, though I'm not sure if he used the approach in that blog post, and...
I think the main benefit to this is the ability to give type annotations on variables, instead of just methods and method arguments. What needs to be done is just...
It's standard to just ignore masking when using a CNN on word / character sequences, because the max pooling can effectively ignore the padding tokens, anyway. It'd be interesting to...
You could make the argument that the way data is handled and the way we build models are too tightly coupled, and should be decomposed. That would mean, basically, making...
We have a few places under `doc/` where we've written explanations of things. Those explanations really should be in READMEs in the code, and just _read_ from the README and...
Similar to what is done in this paper: https://www.semanticscholar.org/paper/Multi-Perspective-Context-Matching-for-Machine-Wang-Mi/e94697b98b707f557436e025bdc8498fa261d3bc.