Chatbot Process - YooEunseok/Graduation_Project_GARVIS GitHub Wiki

Overall Process

The overall process structure of the chatbot system is shown below.

We will explain the contents of the schedule correction function among various functions as an example.

1. Intent Matching

The chatbot determines that it intends to modify the schedule after seeing the utterances of the above users.

2. Entity Matching

Gets the entity 'Date', 'What you want to modify', and 'Schedule Name' required to modify the schedule in the user's utterance sentence.

3. Identify detailed entity types

At this point, the original date of the schedule was implemented so that it could be possible for both a specific date, such as today, or a period, such as this week. Also, we have implemented the functions that you want to modify by dividing the time only, the date only, or the time and the date all into three categories.

4. Database Searching

Schedules corresponding to a single data or period are read from the database in turn

-> Count how many schedules are there that contain a user-schedule entity

-> More than one: Ask the user to choose 1: Determining it by the schedule 0: The string similarity test determines the most similar schedule

5. Execute function

Finally, when the user receives a choice or a positive response, the data in the database for the corresponding schedule is modified to the user's desired content.

In addition, sentences that ignore small typos or spaces are also well matched.