Sei sulla pagina 1di 3

OVERVIEW

Introduction ! Suggest Edits (/docs/key-concepts/edit)

KEY CONCEPTS
Understand the key concepts
Introduction (/d
(/d……
Agents (/docs/conc…
Machine Learning (/… API.AI: NLU and Dialog Management
Platform Integrations
Entities (/docs/con…
Input Methods
Intents (/docs/conc…
Output Methods
Actions and Param… FulSllment
Rich Messages (/do… Get Started Today
Response design fo…
Contexts (/docs/co…
Events (/docs/conc… API.AI is a platform for building conversational interfaces for bots,
Dialogs (/docs/dial… applications, and devices.
Training (/docs/trai… The diagram below shows how API.AI is related to other components
Machine Learning S… and how it processes data:
Prebuilt Agents (/do…
Analytics (/docs/an…

API REFERENCE

FULFILLMENT

SDKS

INTEGRATIONS
(https://Sles.readme.io/a769bab-API-AI_key-concepts.png)
DOMAINS (DEPRECATED)
In the diagram, the green is provided by the API.AI platform. Your app /
GUIDELINES
bot / device code provides the input and output methods and responds
VIDEOS to actionable data. You can also provide an optional webhook
implementation which API.AI uses to connect to your web service. Your
web service can then perform business logic, call external APIs, or
access data stores.

API.AI: NLU and Dialog Management


API.AI receives a query as input data. A query
(https://docs.api.ai/docs/query) is either text in natural language or an
event (https://docs.api.ai/docs/concept-events) name sent to API.AI.

API.AI matches the query to the most suitable intent


(https://docs.api.ai/docs/concept-intents) based on information
contained in the intent (examples, entities
(https://docs.api.ai/docs/concept-entities) used for annotations
(https://docs.api.ai/docs/concept-intents#section-example-annotation),
contexts (https://docs.api.ai/docs/concept-contexts), parameters
(https://docs.api.ai/docs/concept-actions), events) and the agent
(https://docs.api.ai/docs/concept-agents)'s machine learning
(https://docs.api.ai/docs/machine-learning) model. API.AI transforms
the query text into actionable data and returns output data as a JSON
response object (https://docs.api.ai/docs/query#response).

The process of transforming natural language into actionable data is


called Natural Language Understanding (NLU). Dialog management
tools such as contexts (https://docs.api.ai/docs/concept-contexts) and
intent priorities (https://docs.api.ai/docs/concept-intents#intents-
priority) allow developers to control the conversation _ow.

Platform Integrations

There are several ways of connecting an API.AI agent to your app, bot, or
device:

You can send GET and POST HTTP requests to your agent using
the /query endpoint (https://docs.api.ai/docs/query).
You can use our SDKs (https://docs.api.ai/docs/sdks) to send
query requests to different platforms.
For popular messaging platforms, you can use one-click
integrations (https://docs.api.ai/docs/integrations) that allow you
to create bots without any coding. You can also use our source
code from Github (https://github.com/api-ai) to extend these
integrations and create custom bot implementations.

Input Methods
API.AI is designed to receive text as input data. If you want to send voice
requests to your app, bot, or device, consider using Google Speech API
(https://cloud.google.com/speech/) or other 3rd party solutions which
can convert speech to text.

Depending on the app, bot, or device platform, speech recognition might


be available natively. For example, if you are building an Action on
Google (https://docs.api.ai/docs/actions-on-google-integration), an
Android or iOS app, speech recognition is supported natively.

Output Methods

API.AI allows you to deSne text responses


(https://docs.api.ai/docs/concept-intents#section-text-response) for
your app, bot, or device directly in intents. For some messaging
platforms, you can also deSne rich messages
(https://docs.api.ai/docs/rich-messages) as your bot responses.

If you want your app, bot, or device to provide voice responses, you can
integrate your agent with a 3rd-party text-to-speech engine.

FulEllment

Optionally, you may want your app, bot, or device to perform operations
(e.g., open applications, extract data, do math, etc.) and provide
responses to those operations. We use the term fulEllment to refer to
such operations and responses.

FulSllment can be implemented in your app, bot, or device code.

You can consider creating a web service and integrating it with your
API.AI agent via an API.AI webhook
(https://docs.api.ai/docs/webhook). In your web service, you can make
calls to other APIs and send results back to your API.AI agent as
responses.

Get Started Today

Potrebbero piacerti anche