Mindmeld - An Open-source, Deep-domain Conversational AI Toolkit

Explore an open-source platform for building advanced enterprise conversational AI assistants in this 30-minute talk. Learn about Cisco's Mindmeld toolkit, which supports deep-domain knowledge bases for complex use cases. Discover the challenges of creating production-quality conversational interfaces and why current machine learning tools fall short. Delve into the components of the Mindmeld Conversational AI Platform, including the Natural Language Processor, Domain Classifier, Intent Classifier, Entity Recognizer, Role Classifier, Entity Resolver, Language Parser, Dialogue Manager, and Application Manager. Examine a sample interaction with an enterprise voice assistant, covering domain classification, intent classification, entity recognition, role classification, language parsing, dialogue management, question answering, and entity resolution.

Intro
Conversational interfaces are the next frontier
Building production-quality conversational assistants is one of today's hardest Al challenges
Today's machine learning tools are insufficient Cloud NLP Services
The Mind Meld Conversational Al Platform
Github
Natural Language Processor understands the user query
Domain Classifier assigns the user query to a pre- defined domain
Intent Classifier determines the user intent
Entity Recognizer detects all relevant entities in the user query
Role Classifier assigns role labels to extracted entities
Entity Resolver transforms each extracted entity into its canonical form
Language Parser clusters extracted entities into meaningful entity groups
Dialogue Manager directs the flow of the conversation
Application Manager is the platform orchestrator
A sample interaction with an enterprise voice assistant
Domain Classification
Intent Classification
Entity Recognition
Role Classification
Language Parsing
Dialogue Management & Question Answering
Entity Resolution