Whizbot

Engage into interesting conversations with this multi-purpose chatbot

Description

This project interesting conversational agent, beginning with:

• describing itself and its current abilities;
• answering open-ended questions with information from Wikipedia;
• showing Weather data, optionally specifying location; if none is given, the user is asked for the desired city;
• telling Chuck Norris jokes, optionally specifying the category; the list of categories can also be requested.

Implementation

Frontend

• languages: HTML5, JavaScript (ES6), CSS3;
• Rivets.js - for data-binding;
• DOMtastic - for DOM manipulation (and traversal).

Backend

Python3 ASGI Web server hosted on the CloudFoundry service.

• Uvicorn;
• BareASGI - for serving static files and handling HTTP requests;
• Watson Assistant, for classifying natural language intents ;
• Wikipedia Search API - for retrieving answers to open-ended questions;
• Weather Company Data for IBM Bluemix - for fetching weather information;
• Chuck Norris jokes API.

Details

Watson Assistant

This project is utilizing two Assistant instances:

• chatbot - for regular intent processing;
• wikibot - for storing search snippets obtained from Wikipedia as intents (questions) with corresponding dialog nodes (answers).

The current model supports the following types of inquiries:

• chatbot-related: identity, description, capabilities;
• weather-related: general information, temperature, wind speed, optionally specifying location; uses last-specified location if none is given;
• Chuck Norris jokes;

In the case where no specific intent is recognized, the query is interpreted as an open-ended question and handled via Wikipedia's Search API; the result is stored in wikibot for later retrieval.

Business Model Canvas