Universal Sentence Encoder

There’s a great post by Georg Weise at Scalable Minds on using the Google Sentence Encoder with Rasa. I’ve implemented the pipeline as he describes and I’d like to add entity extraction.

I’ve tried the following pipeline but I’m still not having any success extracting entities. Any suggestions on getting a pipeline along these lines working?

language: "en"

- name: "tokenizer_whitespace"
- name: "ner_crf"
- name: "ner_synonyms"
- name: "intent_featurizer_count_vectors"
- name: "use_featurizer.UniversalSentenceEncoderFeaturizer"
- name: "intent_classifier_tensorflow_embedding"

I’m now running a separate NLU docker image and in the process, entity extraction with TensorFlow is now working.

Hi @stephens, did you push your model into GitHub? If so could you share the link?

Thanks Serdar

Here’s a gist