mosen/starter-pack-rasa-nlu

Rasa NLU starter-pack

Looked through the Rasa NLU documentation and ready to build your first intelligent assistant? We have some resources to help you get started! This repository contains the foundations of your first custom assistant. This starter-pack is also accompanied by a step-by-step video tutorial, which you can find here.

This starter-pack comes with a small amount of training data which lets you build a simple assistant. You can find more training data here in the forum and use it to teach your assistant new skills and make it more engaging.

We would recommend downloading this before getting started, although the tutorial will also work with just the data in this repo.

The initial version of this starter-pack lets you build a basic Rasa NLU model capable of understanding few simple intents and entities.

Clone this repo to get started:

git clone https://github.com/RasaHQ/starter-pack-rasa-nlu.git

After you clone the repository, a directory called starter-pack-rasa-nlu will be downloaded to your local machine. It contains all the files of this repo and you should refer to this directory as your 'project directory'.

Setup and installation

If you haven’t installed Rasa NLU yet, you can do it by navigating to the project directory and running:

pip install -r requirements.txt

You also need to install a spaCy English language model. You can install it by running:

python -m spacy download en

What’s in this starter-pack?

This starter-pack contains some training data and the main files which you can use as the basis of your first custom assistant. It also has the usual file structure of the assistant built with Rasa Stack. This starter-pack consists of the following files:

• data/nlu_data.json file contains the training examples of five simple intents:

  • greet
  • goodbye
  • thanks
  • affirm
  • name (examples of this intent contain an entity called name)

• nlu_cofing.yml file contains the configuration of the Rasa NLU training pipeline:

language: "en"

pipeline: spacy_sklearn

How to use it?

• NOTE: If running on Windows, you will either have to install make or copy the following commands from the Makefile Even though this starter-pack have only five and quite generic intents, you can train the Rasa NLU model by running:
  make train-nlu
  This will train the NLU model and store it inside the /models/current/nlu folder of your project directory.

To test the model, you can run it as a server using the following command which will start the server using port 5000:
make run-nlu

To get the results of the model, you can pass an input message by making a request:
curl XPOST localhost:5000/parse -d '{"query":"Hello", "project": "current"}'

What's next?

Five intents and one entity are definitely not enough to build an awesome assistant so here are some ideas for what you can do to take this project to the next level:

• Use the Rasa NLU training data file which you downloaded previously from Rasa Community Forum. This dataset contains quite a few interesting intents which will enable your assistant to handle small talk. To use it, append the training examples to data/nlu_data.md file, retrain the NLU model and see how your assistant learns new skills.
• Enrich the data/nlu_data.md file with the intents you would like your bot to understand. Retrain the NLU model using the command above and see you assistant improving with every run!

Make sure to let us know how you are getting on and what have you built. Visit Rasa Community Forum and share your experience.