After augmenting the dictionary of Jieba, do I need to re-train the NLU model?

Rasa Open Source
1

I have configured the JiebaTokenizer as follows:

  • name: JiebaTokenizer dictionary_path: “./jieba_userdict” “intent_tokenization_flag”: True “intent_split_symbol”: “+”

and a dictionary is put in the dictionary_path.

However, I found an identical file in the model path, ./models/nlu/component_1_JiebaTokenizer/

The question is, after I modified the Jieba dictionary file, do I need to re-train the NLU model every time or just copy the dictionary file to the model path “./models/nlu/component_1_JiebaTokenizer/”?