Skip to content

Latest commit

 

History

History
91 lines (65 loc) · 3.09 KB

README.md

File metadata and controls

91 lines (65 loc) · 3.09 KB

Assist.Jarvis

(Proof Of Concept)

A functional PoC chatbot inspired from Google's DialogFlow

Google's DialogFlow

DialogFlow is a framework by google that allow developers to create their own goal oriented chatbot using Machine Learning and NLP.

* The pre-trained model is hypertuned to get some specific results, train your own model to get better results...

About This Project

How it works

It primarily functions in 4 different phase:

(1) It classifies the Intent using an LSTM Intent Classifier.
(2) It parses all the Entities from the given text input.
(3) The Entities are then passed to an Action(a function) associated with the Intent.
(4) A Sequence-to-Sequence model is used to generate the final output response.


Internal operation for a weather query

Dataflow for the weather intent :




Intent Classifier

Our model uses a simple 3 Layered Intent Classifier.

The First layer is an LSTM followes by 2 Dense layers.

LSTM Layer

Entity Parser

The Entity Parser uses Spacy library to do NLP tasks

Actions

Functions to parse weather, do a search, etc....

Dialog Model

The Dialog Generation is done by a Sequence-to-Sequence Model.


Usage

The python files are stored in the './src/' folder. A standalone jupyter notebook is saved inside the folder notebook. The Folder imgs contain screenshots of the program.

Run the file 'main.py' inside the folder './src' to run the program.

* Estimated time for the application to load : 36 seconds

Try typing Hi or hello.....

First, it shows the intent of the text, then the entities, the generated dialog and at last, the reply from the bot...

Try asking about the weather.....