Intelligent Bots as Service Proxies

Florian Rappl, iQuest

Intelligent Bots

Service Proxies for the Next Generation

Cover Bots

Florian Rappl

Solution Architect

  • IoT / Embedded Computing
  • .NET Communication Systems
  • Web Technologies

Writer and consultant

  • Microsoft MVP for C# / Development Tools
  • Active contributions to open-source projects
  • Articles for CodeProject, SitePoint, tuts+, ...
S1M0NE Movie
Surrogates Movie
Microsoft Twitter Bot Tay.ai

Agenda

  1. Service Proxies
  2. Ingredients for an Intelligent Bot
  3. Demo Session

Service Proxies

Problem Situation

Service Proxy Situation

Problem Situation

Service Proxy Breaking Change

Problem Situation

Service Proxy Breaking Change

Problem Situation

Service Proxy Breaking Change

That reminds me ...

XKCD Standards

Ideal Design

Service Proxy Ideal Case

Problem Solution

Service Proxy Generic Protocol

Realistic Implementation

Service Proxy Implementation

Example Illustration

Service Proxy Example

Demo

Reasoning

  • Base level for
    • advanced services,
    • higher machines, and
    • humans
  • Language is interoperable and extensible
  • Centers around humans instead of machines
  • Naturally includes digital assistents

Ingredients

Application Layers

Bot Architecture Layers

The Microsoft Bot Framework

  • A framework and service consisting of three parts
    • Bot builder to build bots (producers)
    • Bot connector to connect bots (distributors)
    • Bot directory to discover bots (consumers)

The Microsoft Bot Framework

Bot Framework Layers

The Microsoft Bot Framework

  • Persistence is added via the connector
  • Simple key-value pairs
  • Allows to retrieve context information per user and/or conversation
  • Storing meaningful information for the conversation is key

The Microsoft Bot Framework

Bot Framework Teaser

For each channel we need to sign up at the used service provider

Example: Connecting Skype

Connecting Skype Bot Framework

The Microsoft Bot Framework

  • Stateless communication with the connector
  • Theoretically limitless scalability
  • A single bot can be sustained by multiple instances

Demo

LUIS - Introduction

LUIS Bot Framework
  • Language Understanding Intelligence Service
  • Part of project Oxford (now Cognitive Services)
  • Text recognition with entity identification
  • Provide intents and entities
  • Train new utterances to improve accuracy
  • Publish API hosted by Microsoft

LUIS - Usage

LUIS Usage

Demo

Thanks for your attention

  • Feel free to contact me