PRIYANKA VERGADIA: Welcome
to "Deconstructing Chatbots." I am Priyanka Vergadia. And in this episode, we will
dissect the architecture of chatbots to better
understand the building blocks. [MUSIC PLAYING] In the last episode, we
introduced Dialogflow, which is an end-to-end
tool powered by natural language
understanding to facilitate rich and natural conversations. Today let's look at a
high-level architecture. Dialogflow sits in the
middle of the stack. A user can interface with it
via all the common channels, including text, websites, apps,
messengers, and smart voice devices like Google Home. Dialogflow handles the job of
translating natural language into machine-readable data
using machine learning model trained by your examples. Once it identifies what
the user's talking about, it can hand this
data to your back-end where you can use it
to make stuff happen. At the back-end you
can fulfill the request by integrating with your other
services, databases, or even third party tools like your CRM. Now let's dig one layer deeper
into the Dialogflow piece itself. We first create an
agent within Dialogflow. Agent is basically your
entire chatbot application-- the experience of collecting
what the user's saying, mapping it to an intent,
taking an action on it, and then providing the
user with the response. And in your agent, this all
starts with the trigger event called utterance. This is how our users
invoke the chatbot. So if I say, hey
Google, play some music, the whole sentence is utterance. While the phrase, hey
Google, is the trigger. Let's take another example. Hey Google, talk
to Smart Scheduler. The phrase, talk
to Smart Scheduler, is the invocation
phrase for our chatbot. And Smart Scheduler here
is the invocation name. Once the bot is activated
and has collected the user utterance, we need to understand
what the user's intent is. Why do they want
to talk to our bot? So when you say, I want
to set an appointment, set appointment
here is the intent. Or if you ask what are
your hours of operation, then hours of operation
is the intent. To control all this,
you provide Dialogflow with different examples
of users' intent like set an appointment, hours
of operation, and others. Dialogflow then trains
the machine learning model with many more similar
phrases and finally maps the user's phrase
to the right intent. This process is called
intent matching. Now that we know
our user's intent we need to know what to do
to give them a response. For that you configure
actions and parameters to define the variables you
want to collect and store. Let's look at an example. Set an appointment
for 5:00 AM tomorrow. When a users says that,
5:00 AM and tomorrow are two critical pieces of
information in that statement that we would actually need
to book an appointment. Those variables are
defined as entities. Dialogflow offers different
types of entities. And we will cover those in
detail in an upcoming episode. Once we have the
variables, we may use them to provide a
static response to the user. Or in most cases, we may
want to send the variable to our back-end, take
some action on it, and then provide a user
with a dynamic response. We will look at
that in a second. OK. To summarize, an intent includes
training phrases, actions and parameters, and response. Depending on what
services your bot offers, you might typically have from
few to thousands of intents. They could also be in
different languages. While you're looking
at the architecture, it's worth mentioning context. Context is the method for your
chatbot to store and access variables so it can exchange
information from one intent to another in a conversation. We will have more information on
context in an upcoming episode. Fulfillment is the last
piece in the puzzle. It's the code that
you write to interface with the back-end services to
respond to a dynamic request. We will play with
this a lot more soon. But for now, note
that Dialogflow has inbuilt integration
with Google Cloud Functions to interface with your back-end. And you can also provide
an other HTTPS endpoint, and Dialogflow will
just connect to it. All right. So today we looked at a
simple high-level architecture of setting up an agent
within Dialogflow and learned about intents,
entities, context, and fulfillment at a high level. Next we are going to go hands-on
and build a simple appointment scheduler chatbot. If that sounds exciting
to you, join me in the next episode of
"Deconstructing Chatbots." [MUSIC PLAYING]
