Trouble extracting entities

rasa-nlu

(Truong-Allié) #1

Hello, I am new to rasa, and I have troubles training my nlu model… Here is my problem : When I have only one entity in my sentence, it is ok, and the model extracts it, but when I have two entities, it extracts only one of the two entities.

For instance, if I have the sentences : « I want to go to a stop ». It does extract « stop » as info_type, « I am looking for Bastille », it does extract « Bastille » as stop_name But when I write : « I want to go to the stop named Bastille ». I can either have « Bastille » extracted, either « stop », but not both of them.

Could someone help me please ?

My spacy configuration is the following :

pipeline:

  • name: "nlp_spacy"
  • name: "tokenizer_spacy"
  • name: "intent_featurizer_spacy"
  • name: "intent_classifier_sklearn"
  • name: "ner_crf"
  • name: “ner_synonyms” path: “./models/nlu” data: “./data/data.json”

language: “fr”

Thanks !


(Akela Drissner) #2

You need to add training examples like that


(Truong-Allié) #3

Thank you, it seems to work ! I thought I had enough examples, but I was wrong. Now I have about 20 examples for each type of sentence, and it seems ok