Skip to content

Latest commit

 

History

History
executable file
·
78 lines (52 loc) · 2.22 KB

README.md

File metadata and controls

executable file
·
78 lines (52 loc) · 2.22 KB

Development of Dialog System within Rasa in the Movie Domain

This is the final project of the Language Understanding System course, the aim is to develop o Dialog System within Rasa in the Movie Domain capable of understanding what the user requires and of providing the correct answer.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

To succesfully run the project you need to download two external tool:

  • Rasa-Core, see RasaCore

  • Rasa-NLU, see RasaNLU

  • Rasa additional dependecies:

    • Spacy, see Spacy
    • CRF Suite, through pip
    pip install sklearn_crfsuite

Make the project run

To run the system, the procedure is very straight:

  • From terminal, move to the /scripts folder
cd scripts
  • Choose which options you want to perform between training the NLU, training the dialogue manager and run and launch the script with the respective parameter
python bot.py <parameter>

with from train-nlu, train-dialogue, run

  • The python script will perform all requested action

Deployment

  1. train-nlu
    First action to be performed. Main operations are:

    • Load the training data
    • Pass the domain file (.yml) to the Trainer
    • Train the NLU
    • Results are stored in:
    /models/default/chat

  2. train-dialogue To be performed immediately after the training of the NLU, main operations are:

    • Create the RasaNLUInterpreter with the result of the previous training as input
    • Create the Agent and setting its caracteristics: Policies (Memoization, Keras), Interpreter, configuration file of the NLU
    • Train the Agent with the stories, namely a file of exaples of conversations
    • (Optional) Enable the "Training online", that allows you to manually correct bot intents/actions to improve its precision
    • Results are stored in:
    /models/dialogue

  3. run

    • Load the interpreter
    • Pass the interpreter to the agent
    • Run the bot

Authors

  • Andrea Montagner - ID:189514