Adding a nlu engine

The NodeJS skill boilerplate and NodeJS SDK support the regexp and the wcs (Watson Assistant) nlu engines out-of-the-box. You can add other nlu engines to your skill.

An example of a custom nlu engine is provided in the ../res/nlu/example folder in the skill NodeJS boilerplate. The ../res/nlu/example/nlu.js file controls the functioning of the nlu engine. The ../res/nlu/example.json file provides external options to the nlu engine.

Procedure

Complete these steps:

1. In your custom skill, create a folder for the nlu engine in ../res/nlu. Use a folder name that represents the nlu engine, for example, rasa.
2. Copy the file ../res/nlu/example/nlu.js to ../res/nlu/your_custom_nlu_folder.
3. Edit the nlu.js file. The file provides the following functions:
   • A constructor for the nlu engine.
   • An init function that initializes the nlu engine.
   • An evaluate function that you customize to extract entities and intents. The function returns entities and intents with confidence scores in an intentities object. Follow the instructions in the file.
4. Create a your_custom_nlu_engine.json file in ../res/nlu/. Add any external information to the file that your nlu engine requires, for example, credentials. Tip: The file is mandatory but can contain an empty object only.
5. Add the custom nlu engine to the list of nlu engines in the manifest file. For example:

      "nlu": [
          "wcs", "regexp", "skill", "rasa"
      ]
   

Next topic

Configuring skill authentication