APPS ON GOOGLE ASSISTANT USING ACTIONS GAURAV BHATNAGAR 

AGENDA • Background • Google Assistant Workflow • How Assistant provides customized response. • Dialogflow and all its main components • Demo of an example • Basics of Conversation design • GUI vs VUI 

BACKGROUND • Google Assistant is currently available on more than 400 million devices around the world. • It is available in phones(iOS or android), tablets, cars, Chromebook & Home Devices. 

GOOGLE ASSISTANT STATS 

GOOGLE ASSISTANT WORKFLOW Source : https://dialogflow.com/docs/agents 

DIFFERENT APPROACHES FOR CUSTOMIZED ASSISTANT APPS 

GOOGLE ASSISTANT WORKFLOW (USING DIALOGFLOW) Source : https://dialogflow.com/docs/agents 

DIALOGFLOW • Agents Agents are NLU Modules built with the help of contexts, intent priorities, responsibilities, and fulfillment via webhook. • Intents They represent a mapping between what a user says and what action should be taken by your customized software. • Contexts They are designed for passing on information based on user’s preferences, geographic location. • Fulfillment This component connects your Dialogflow agent to your services, APIs and/or databases. It connects to webhooks via Actions on Google. • Entities, Actions, Parameters….. 

INTEGRATIONS PROVIDED BY DIALOGFLOW Source : https://dialogflow.com/docs/integrations 

SHOWING AN EXAMPLE DEMO 

CONVERSATION DESIGN • It is basically a design language based on human conversation. • We are providing an experience and merely not just an application. • Creating a System and User persona. • Conversation Design should be built on cooperative principle. Conversations should always be designed to provide comfortable, frictionless and user centric experience. • Group devices by the components used in conversation : For e.g. Google Home/ Pixel. 

BUILDING BLOCKS FOR CONVERSATION • Expect users to give more information. • Try to always keep the dialog towards the use case you are trying to solve. • Try to always move the conversation forward. • Try to keep the conversation optimized for relevance and provide clarity in answers. • Try to understand the context and listen between the lines. • Follow-up conversations should have the context attached. • Keep the conversation going by taking turns and not try to monopolize. 

No content

ALWAYS KEEPING REAL LIFE USER CONDITIONS IN MIND Interrupted Cut off too soon No way out Accent Language understanding Self Correction Confused Side Speech Valid but not in grammar Didn’t Understand Background noise Ask a question back. Coughs Talked too long Too quiet Hesitate Know options by some other manner Stutter Start Too many choices Poor connection In Conversations, there are no “Errors”. 

REFERENCES • Actions on Google Developers : https://developers.google.com/actions. • https://developers.google.com/actions/design/ • https://dialogflow.com • https://caster.io/courses/google-assistant-app-with-zero-code