Rasa_Japanese

Rasa is an open source machine learning framework to automate text-and voice-based conversations.Rasa's primary purpose is to help you build contextual, layered conversations with lots of back-and-forth. To have a real conversation, you need to have some memory and build on things that were said earlier. Rasa lets you do that in a scalable way.

How to use rasa for japanese language?

Installation

• Firstly install rasa in your own environment as following:

git clone https://github.com/RasaHQ/rasa.git
cd rasa

• As we will add a custom component to tokenize japanese text , we will use mecab library for tokinzing the text.
  add mecab-python3 at the end of requirements.txt file. Or install mecab using pip.

pip install mecab-python3

• Now install all the requirements using :

pip install -r requirements.txt
pip install -e .

Create Project

• First run the following command:

   rasa init --no-prompt
   
  

  if you find any error like " rasa.core.trackers - Tried to set non existent slot 'name'. Make sure you added all your slots to your domain file.", then please remove all the data inside rasa/data/* directory.

• Then add the Japanese language tokenizer in the path "rasa/rasa/nlu/japanese_tokenizer.py". I have added the file in "rasa/rasa/nlu/japanese_tokenizer.py" path.

• Add JapaneseTokenizer component class in /rasa/rasa/nlu/registry.py. E.g:

  from rasa.nlu.tokenizers.japanese_tokenizer import JapaneseTokenizer
  
  

  Add class name inside component_classes = [] dictionary. E.g

  component_classes = [
  # tokenizers
  JapaneseTokenizer,
  MitieTokenizer,
  SpacyTokenizer,
  WhitespaceTokenizer,
  JiebaTokenizer,
  ]
  

• Now add JapaneseTokenizer as pipeline in config.yml as follows:

   # Configuration for Rasa NLU.
   # https://rasa.com/docs/rasa/nlu/components/
   language: jp
   pipeline:
     - name: "JapaneseTokenizer"
     - name: "RegexFeaturizer"
     - name: "CRFEntityExtractor"
     - name: "EntitySynonymMapper"
     - name: "CountVectorsFeaturizer"
     - name: "EmbeddingIntentClassifier"
   # Configuration for Rasa Core.
   # https://rasa.com/docs/rasa/core/policies/
   policies:
     - name: MemoizationPolicy
     - name: KerasPolicy
     - name: MappingPolicy
  
  

• Now Add data as data/nlu.md, stories.md and domail.yml. E.g:

  nle.md:

  ## intent:greet
  - ハロー
  - もしもし
  - 初めまして
  - こんにちは
  - はじめまして
  
  ## intent:icebreak12
  - よろしくお願いします
  - こちらこそ
  - 宜しくお願いします
  - 問題ないです
  

  stories.md:

   ## happy path
   * greet
     - utter_greet
   * icebreak12
     - utter_icebreak12
  

  domain.yml:

  intents:
  - utter_icebreak12
  - greet
  templates:
    utter_icebreak12:
    - text: 今どこで、何をしていますか？
    utter_greet:
    - text: ご協力いただきありがとうございます。本日は宜しくお願いします
  actions:
  - utter_greet
  - utter_icebreak12
  

• Now train your model using:

    rasa train
  

• To chat run:

    rasa shell