1. NLU (Natural Language Understanding)
The purpose of NLU is to complete the business email list understanding of user instructions.
Therefore, the input of the NLU module is a user command, and the output mainly adopts the representation structure of DIS, D stands for Domain (domain), I stands for Intent (intent), and S stands for Slot (slot).
(In some cases, there is no Domain, and the direct output is Intent)
Domain, Intent, and Slot are all predefined by product managers according to their needs. Different business scenarios will define different DIS.
For this example of booking a meeting room, DIS can be defined as follows:
2. DM (Dialog Management)
Dialogue management DM controls the process of human-machine dialogue and is the core of ChatBot.
During the dialogue process, the dialogue robot system will continuously decide the optimal action to be taken in the next step according to the current dialogue state and user behavior, so as to complete the entire dialogue task.
Therefore, in this process, DM mainly completes the following two tasks:
(1) Maintain & update dialog state (dialog state tracking, DST)
Maintain an up-to-date state of the conversation based on previous states and user input.
(2) Generate a system decision (dialog strategy)
System decisions are made based on the dialog state in DST to decide what to do next.