Custom Alexa Skills
by: Deepansh Sachdeva
Voice Assistants are COOL... but NOT EVERY TIME
Types of Skills
➔ Custom Skills
➔ Smart Home Skills
➔ Video Skills
➔ Flash Briefing Skills
➔ Music Skills
Custom Skills
➔ A skill that can handle just about any type of request. Just about anything else
you can think of from interactive games to integrating web services.
➔ Components of custom skills
◆ Intent: actions that users can do with your skill.
◆ Utterances: words and phrases users can say to invoke those intents.
◆ Invocation Name: used to identify the skill and start the conversation.
◆ Cloud Service: endpoint that handles the intent requests.
User-Skill Communication
Steps to Build a Custom Skill
➔ Setup skill in developer console
➔ Build interaction model
➔ Code & Deploy your endpoint service (AWS Lambda recommended)
➔ Beta testing (optional)
➔ Submit for certification
Interaction Model
➔ The interaction model refers to your collection of intents, sample utterances,
and the dialog model.
➔ The requests your skill can handle are represented as intents.
➔ Intents can optionally have arguments called slots.
➔ Sample utterances map the intents to the words and phrases users can say to
interact with your skill.
➔ A dialog model identifies information your skill requires and the prompts Alexa
can use to collect and confirm that information in a conversation with the user.
Intents & Slots
➔ Intents can be custom with specified utterances or amazon provides few
built-in intents.
➔ Each intent required at least one sample utterance.
➔ Intents have slots to capture data from speech. It can also be custom and
built-in.
➔ Dialog model can be used for multi turn conversation.
Dialog Model for Multi Turn Conversation
➔ A dialog model identifies the prompts and user utterances to collect, validate,
and confirm slot values and intents.
➔ Determines the next step in conversation & ask the user for more information.
➔ Components of dialog model:
◆ Required slots & prompts
◆ Slot confirmation prompts
◆ Intent confirmation prompts
◆ Slot validations & prompts
➔ Intent requests provides dialog state which can be STARTED, IN_PROGRESS
or COMPLETED.
Endpoint (AWS Lambda)
➔ The ASK SDK for Node.js is available for easy implementation of lambda for
Alexa skills.
➔ Process intent requests.
➔ Defines intent handlers.
➔ Builds speech responses & directives.
➔ Manages session & persistent attributes.
➔ Alexa service APIs.

Building Alexa Skills

  • 1.
    Custom Alexa Skills by:Deepansh Sachdeva
  • 2.
    Voice Assistants areCOOL... but NOT EVERY TIME
  • 3.
    Types of Skills ➔Custom Skills ➔ Smart Home Skills ➔ Video Skills ➔ Flash Briefing Skills ➔ Music Skills
  • 4.
    Custom Skills ➔ Askill that can handle just about any type of request. Just about anything else you can think of from interactive games to integrating web services. ➔ Components of custom skills ◆ Intent: actions that users can do with your skill. ◆ Utterances: words and phrases users can say to invoke those intents. ◆ Invocation Name: used to identify the skill and start the conversation. ◆ Cloud Service: endpoint that handles the intent requests.
  • 5.
  • 6.
    Steps to Builda Custom Skill ➔ Setup skill in developer console ➔ Build interaction model ➔ Code & Deploy your endpoint service (AWS Lambda recommended) ➔ Beta testing (optional) ➔ Submit for certification
  • 7.
    Interaction Model ➔ Theinteraction model refers to your collection of intents, sample utterances, and the dialog model. ➔ The requests your skill can handle are represented as intents. ➔ Intents can optionally have arguments called slots. ➔ Sample utterances map the intents to the words and phrases users can say to interact with your skill. ➔ A dialog model identifies information your skill requires and the prompts Alexa can use to collect and confirm that information in a conversation with the user.
  • 8.
    Intents & Slots ➔Intents can be custom with specified utterances or amazon provides few built-in intents. ➔ Each intent required at least one sample utterance. ➔ Intents have slots to capture data from speech. It can also be custom and built-in. ➔ Dialog model can be used for multi turn conversation.
  • 9.
    Dialog Model forMulti Turn Conversation ➔ A dialog model identifies the prompts and user utterances to collect, validate, and confirm slot values and intents. ➔ Determines the next step in conversation & ask the user for more information. ➔ Components of dialog model: ◆ Required slots & prompts ◆ Slot confirmation prompts ◆ Intent confirmation prompts ◆ Slot validations & prompts ➔ Intent requests provides dialog state which can be STARTED, IN_PROGRESS or COMPLETED.
  • 10.
    Endpoint (AWS Lambda) ➔The ASK SDK for Node.js is available for easy implementation of lambda for Alexa skills. ➔ Process intent requests. ➔ Defines intent handlers. ➔ Builds speech responses & directives. ➔ Manages session & persistent attributes. ➔ Alexa service APIs.