Entity extraction in multi-intents?

rasa-nlu

(Zoltan Fedor) #1

I have read the blog piece about the handling of multiple intents (see https://blog.rasa.com/how-to-handle-multiple-intents-per-input-using-rasa-nlu-tensorflow-pipeline/), which seems to be essential to handle real conversations with humans.

There was a comment made in the post about entities are not being (yet) available with multiple intents:

For now, TensorFlow pipeline is only performing intent classification tasks, but some really exciting updates regarding entity recognition are on the way so stay tuned for more tutorials and posts coming from us on this new pipeline!

Do we have any further news on this?

Being able to extract multiple intents efficiently (without writing combined intents for all possible combinations) is essential, but we can’t really use it until entity extraction is not available.

Thanks


(Akela Drissner) #2

entity extraction is available. The standard config of 0.13.1 for tensorflow_embedding includes it


(Zoltan Fedor) #3

Thanks, just to confirm, you meant the entity extraction for tensorflow_embedding which also support multiple intents, right? Thanks


(Akela Drissner) #4

yes


(Zoltan Fedor) #5

Thanks


(Datisto) #6

@akelad

What is exactly meant by this? Does entity recognition now uses tensorflow embedding? I thought algorithm like ner_crf is a independently separate algorithm? Can you elaborate this more?


(Akela Drissner) #7

Yes it’s separate, the point was that it’s spacy independent now. Before ner_crf needed spacy in the pipeline to work.