I’m trying to extract multiple intents using the TF pipeline example however I can not seem to get any entities to extract using ner_crf
.
My version of Rasa is current at 0.15.1
My config file is pretty standard for the multiple intents classification: language: en
pipeline:
- name: tokenizer_whitespace
- name: intent_entity_featurizer_regex
- name: ner_crf
- name: ner_synonyms
- name: intent_featurizer_count_vectors
- name: intent_classifier_tensorflow_embedding
intent_tokenization_flag: true
intent_split_symbol: +
An example of my training data is, I have almost 1000 variations of common_examples
{
"rasa_nlu_data":{
"regex_features":[
{
"name":"coordinal",
"pattern":"[0-3][0-9][0-9]"
}
],
"entity_synonyms":[],
"common_examples":[
{
"text":"turn right heading",
"intent":"changeHdg",
"entities":[
{
"end":10,
"entity":"long",
"start":5,
"value":"right"
}
]
}
]
}
}
So when I try turn right heading 350
I expect to get (right, 350)
as my entities but I get nothing.
Could anyone lend second pair of eyes?