5 key terms to master Dialogflow

1. Agent

An Agent is a fundamental concept in Dialogflow. It refers to an NLU (Natural Language Understanding) module responsible for processing conversations with users.

Simply put, an agent listens, analyzes, and understands user input (text or voice) and converts it into structured data that programming languages can process.

Each chatbot system requires a dedicated agent in Dialogflow to handle conversations.

Dialogflow also provides pre-built agents designed for common use cases, such as news, weather, and translation, saving developers time in training the agent.

An agent consists of Intent, Entity, and Fulfillment and can be integrated with various platforms. Understanding the Agent concept is essential for mastering Dialogflow.

2. Intent

In Dialogflow, an Intent represents the user's intent or purpose. AI designers train the chatbot to recognize different intents, allowing it to determine what users want from the conversation.

Example:

• User: "Will it rain in Hanoi tomorrow?" → Requires a weather forecast agent, with an intent named "Hanoi Weather"
• User: "Where can I buy pizza in Nha Trang?" → Requires a location-finding agent, with an intent named "Find Pizza Place"

Different sentences can have the same intent. For example:

1. "Will it rain in Nha Trang tomorrow?"
2. "Which days in the next 7 days will be sunny in Nha Trang?"

Both queries seek weather forecast information, so only one intent is needed for both.

3. Entity

An Entity represents key parameters extracted from user input.

In each message, users provide valuable information such as location, time, or specific requests. These important details are considered entities, which the agent identifies and processes accordingly.

Example:

• "Will it rain in Hanoi tomorrow?" → Entities: "Hanoi" (location), "tomorrow" (time)
• "Where can I buy pizza in Nha Trang?" → Entities: "Pizza" (food), "Nha Trang" (location)
• "I want to watch a comedy movie tonight." → Entities: "Movie" (content), "Comedy" (genre), "Tonight" (time)

If an entity doesn’t exist in the system, it must be created and named appropriately.

4. Fulfillment

Fulfillment is a crucial feature in Dialogflow that allows the chatbot to perform actions, such as querying a database, calling an API, or updating external information.

For example, in a banking chatbot, if a user asks:
"What is my account balance?"

The chatbot cannot answer without retrieving real-time balance information.

Using fulfillment, the chatbot calls an external banking system API to fetch the account balance and return an accurate response.

5. Training Phrases

Training Phrases are example sentences users might say to express an intent. Each intent in Dialogflow is trained using these phrases, helping the system understand user input.

Example training phrases for ordering pizza:

• "Hey Google, order a pizza for me."
• "Hey Google, I want to order a pizza."
• "I’d like a pizza."

Even though the wording differs, all these phrases represent the same intent: ordering a pizza.

The more training phrases provided, the better the system learns to recognize different variations of user input. However, quality is more important than quantity—too many similar sentences reduce the chatbot's ability to learn diverse speech patterns.