Python Forum

Full Version: Is model built correctly?
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Hi,

I'm new to neural networks and I want to know whether my model requires more data or it is built not correctly.

I build a CNN model with PyTorch for pos tagging. My train dataset contains 500 tagged sentences

print(sents[70])
print(tags[70])

what explorer was nicknamed Iberis Pilot
['WH', 'NOUN', 'AUX', 'VERB', 'NOUN', 'NOUN']  
but after training, my model is not able to predict even an example from the train dataset.

tagger(['what explorer was nicknamed Iberis Pilot'])
tagger = POSTagger(model, char_vocab, UNIQUE_TAGS, 40, 20)
tagger(['what explorer was nicknamed Iberis Pilot'])

[['VERB', 'NOUN', 'AUX', 'CONJ', 'NOUN', 'NOUN']]
and example that is not from train dataset

tagger(['where can i buy acid'])

[['VERB', 'TO', 'PREP', 'CONJ', 'NOUN']]
So, my model is built not correctly or it requires more data?
While 500 sentences is not a huge number, it does seem that the results should be better than that. I would look at the model