pos tagger by lstm implemented on pytorch
pretrained model weights are at file 'model-file'
to train the model execute 'python tagger_train.py'
to test it with corpus.test use 'python tagger_predict.py corpus.test model-file corpus.out'
to check accuracy of the predicted pos use 'python tagger_eval.py corpus.out corpus.answer'
Pretrained weights are trained on 3 epochs and get 91% accuracy on the test set (1 hour of training on the colab gpu) model uses only char embedding, their convolutions, lstm and dense linear layer to get probabilities with log_softmax