Using the same data and changing only the pipeline pretrained_embeddings_spacy
into pretrained_embeddings_convert
, my chatbot is no longer able to recognize entities within the text.
For example for the user’s message “when to start a treatment in london”, this is the result given by Spacy:
Received user message 'when to start a treatment in london' with intent '{'name': 'when_treatment', 'confidence': 0.8914682416279816}' and entities '[{'start': 29, 'end': 35, 'value': 'london', 'entity': 'city', 'confidence': 0.9784073266757529, 'extractor': 'CRFEntityExtractor'}]'
and this is the one given by ConveRT:
Received user message 'when to start a treatment in london' with intent '{'name': 'when_treatment', 'confidence': 0.9905155897140503}' and entities '[]'
Anybody knows why ConveRT is not working for entities?
Thank you, Tiziano