Chatbot Structure - YooEunseok/Graduation_Project_GARVIS GitHub Wiki

Term

First, let me explain two basic terms used in chatbots.

• Intent: Speaker's Intention
• Entity: Components of a sentence

For example, in the sentence 'Give me cheese pizza', the intent is 'pizza order', and the entity is 'cheese pizza'.

---

Intent

We generate the intents according to main features.

Intents related to the schedule:

• Schedule modification and deletion
• Assignment completion and modification
• class cancellation and make-up class.

Intents related to the team project:

• Schedule request

Intents related to the mail function:

• Sending mail

*The pictures below are the composition of the entities used in this project.

Within each intent, more than 20 sentences were trained as follows.

---

Structure

We tried to express the intents by function in a tree format

• Schedule Modification and Deletion

• Class Cancellation and Make-up class

• Assignment Completion and Modification

• Team project Scheduling and Mail Sending

---

Entity

- System Entity

Let me explain the main entities of the chatbot. First, values related to the schedule, such as date and time, used system entities provided by Dialogflow. When a user says a date or time in a natural language form, it converts it to a certain date or time form.

- Custom Entity

Custom entities that users can create themselves include a Schedule title, Assignment title, Subject title, etc. When a user enters a schedule, assignment, subject, etc., each input value is sent to the Firebase function in real time. Through this, we have implemented it so that model can learn the entity whenever there is a user's input, rather than learning all the values in advance.