For Juman++ to be widely usable, we want to have a documented and stable C API and an option to have a dynamically linked library.
That library probably should use -fvisibility=hidden and explicit visibility on exported symbols on Unixes and __declspec(dllimport/dllexport) on Windows.

The minimal API should be:

1. Loading a model using a config file
2. Analyzing a sentence
3. Accessing

As of PyTorch 1.1.0, the inputs to pack_padded_sequence do not need to be sorted by length if enforce_sorted=False:

https://pytorch.org/docs/master/nn.html#torch.nn.utils.rnn.pack_padded_sequence

enforce_sorted (bool, optional) – if True, the input is expected to contain sequences sorted by length in a decreasing order. If False, this condition is not checked. Default: True.

Thanks fo

We can use nlp-compromise as a JS postagger to showcase postagga running on the browser !

In SpaCy one can access the fine-grained UD tags by the tag_ attribute (see documentation). In this model it only repeats the value of pos_.

Is there any chance to get the CoNLL-U style FEATS from the tagged data for Hungarian?

In a sentence with an uppercase word in the middle, if the uppercase word is correctly tagged as a common noun, Talismane will not find its lemma in the glossary, because the common noun entry is in lowercase. Need to make the lemma search case-insensitive.

We will need to choose any suitable logger library for logging.