I can't seem to find a clear answer from the documentation as to what type of Neural network is being used if any when you build your own NER model from scratch.
When I build the a model from scratch the cfg file thats created shows hidden depth = 1 & the below settings. Does that mean its a CNN with 1 layer ?
"cnn_maxout_pieces":3, "hidden_depth":1, "hidden_width":128, "maxout_pieces":2
In this walk through it seems to suggest that NER system is deep but in my cfg it seems to be set to 1 hidden layer https://spacy.io/universe/project/video-spacys-ner-model#___gatsb
"spaCy v2.0's Named Entity Recognition system features a sophisticated word embedding strategy using subword features and "Bloom" embeddings, a deep convolutional neural network with residual connections, and a novel transition-based approach to named entity parsing"
I am not 100% clear on what the convolutional layer does for a NER only scenario, my training data only has entities no dependency parsing or POS
Here the convolutional layer is mentioned but it seems to be just focused on dependency parsing https://spacy.io/models#architecture
"We first predict context-sensitive vectors for each word in the input:" "This convolutional layer is shared between the tagger, parser and NER, and will also be shared by the future neural lemmatizer"