EdwardBurgin / rasa-for-articulate

rasa-for-articulate

This project is an example of extending the articulate rasa component. Using this project, we added new component named custom_entities in the rasa pipeline.

Build docker image

$docker build -t rasa-for-articulate:latest .

Build image of ner-service

To use the custom_entities in the rasa pipeline, you need to build another docker image using- https://github.com/patelrajnath/ner_service

Update docker-compose.yml in articulate repository

Change-

  rasa:
    image: samtecspg/articulate-rasa:1.0.0
    volumes: ["${MODEL_DIR:-./local-storage/rasa/nlu-model}:/app/projects", "${RASA_CONFIG:-./local-storage/rasa/rasa-config.yml}:/app/config.yml", "./local-storage/rasa/logs:/app/logs"]

INTO

  rasa:
    image: rasa-for-articulate:latest
    volumes: ["${MODEL_DIR:-./local-storage/rasa/nlu-model}:/app/projects", "${RASA_CONFIG:-./local-storage/rasa/rasa-config.yml}:/app/config.yml", "./local-storage/rasa/logs:/app/logs"]

And add docker-context for ner-service

ner-service:
    image: ner-service:latest
    container_name: custom-ner
    restart: always
    ports:
      - "9501:9501"
    volumes:
      - ./ner-training:/app/project-ner/training
      - ./ner-training:/app/project-ner/corpus
      - ./ner-training:/app/project-ner/metrics

Add the component in the pipeline

We can enable this component adding the following config in articulate-

  {
    "name": "custom_entities",
    "arch": "default",
    "dropout": 0.1,
    "accumulate_gradient": 1,
    "patience": 100,
    "max_epochs": 0,
    "max_steps": 1000,
    "eval_frequency": 100,
  }