In case you come across a word that is not part of the original GloVe dataset.
We currently return an empty array.

def get_embedding( word ):
    try:
        return vectors[ indexes[ word ] ]
    except KeyError:
        return []

This however should be an array of zeros with a size matching the dataset sizes:

def get_embedding(word):
    try:
        return vectors[indexes[word]]
    except KeyError:
        return np.zeros(50) 

In this case I'm using numpy because I'm running this in a real python env for preprocessing but could be an floatarray size 50 with all zeros instead.

This solves issues like:

ValueError: all the input array dimensions for the concatenation axis must match exactly, but along dimension 1, the array at index 0 has size 0 and the array at index 2 has size 50

Would be great to allow developers to provide their own GloVe vectors:

• https://github.com/stanfordnlp/GloVe#train-word-vectors-on-a-new-corpus

Would also love to know and request documentation how you generated the pkl and h5 files.

On run of the app, the mainActivity crashes on launch giving the following error

 java.lang.RuntimeException: Unable to get provider androidx.startup.InitializationProvider: androidx.startup.StartupException: java.lang.UnsatisfiedLinkError: dlopen failed: library "libcrypto_chaquopy.so" not found