Skip to content

02-try-it-out.md

Latest commit

 

History

History
92 lines (54 loc) · 2.71 KB

02-try-it-out.md

File metadata and controls

92 lines (54 loc) · 2.71 KB

Try it out

to run code here install Rasa Core and NLU

pip install rasa sklearn_crfsuite prompt_toolkit

Here will be build a chatbot asking and sending fun pic to cheer you up.

Teaching bots to understand user inputs using Rasa NLU

Start by teaching assistant to understand messages

  • train NLU model with plaintext inputs to extract structured data

  • the structured data is called intent

  • first need to define use messages as in nlu-messages.md

  • NLU model configuration defines how NLU model will be trained and features from text input will be extracted, in our example a predfined tensorflow_embedding pipeline as in nlu-config.yml

  • train the NLU model using command below, it will call rasa.nlu and save model inside models/current/nlu

python -m rasa_nlu.train -c nlu-config.yml --data nlu-messages.md -o models --fixed_model_name nlu --project current --verbose
  • test the model to check if bot can understand, check-nlu.py

Teaching the bot to respond using Rasa Core

Define Stories

  • teach your chatbot to respond using Rasa core, it will train dialogue management model and predict how bot should respond at any state

  • learns from real conversational data in form of training stories

  • story is real conversation between user and bot, with user inputs expressed as intent and response of bot as action names
  • sample simple story

  • starts with ## followed by a name

  • lines starting with * are messages from user, not the actual message but the intent that represents what user means

  • lines with - at start are actions by bot, sent to user

  • in general an action can do anything, like calling an API

  • example stories at eg-stories.md

## story1
* greet
  - utter_greet

Define Domain

  • Domain defines vocab of bot

the user inputs expected, what actions to predict, how to respond and what info to store

example domain at eg-domain.yml

  • intent are things you expect user to say

  • actions are what bot can do and say, simple actions start with utter_ demanding a response from template; custom actions for advanced flows

  • templates are baseline for things bot can say

Training dialogue model

  • to train a neural net on example stories, run below command generating trained model at models/dialogue
python -m rasa_core.train -d eg-domain.yml -s eg-stories.md -o models/dialogue

Talk to your Bot

  • to start full bot with Rasa Core and Rasa NLU models using eg-bot.py

updated the chatbot pythonn code to run at cli as REPL instead of in Notebook