AI Understander
AI Understander understands Natural Language statement and identifies Intents & Entities. A confidence score is also calculated along with the matching.
This service can be consumed using the Morph Platform or can be externally accessed using the Understander API.
Intent
Intent helps in understanding appropriate action to be taken by bot depending on the user statement. For example the appropriate intent for Find me red denim jeans can be search.
Entity
Entity are words/phrases detected from user statement which helps bot to apply intent's action on particular data subset or to be used as filter. For example entities in Find me red denim jeans can be red[Color], denim[Type] and jeans[Product].
There are two types of entities you can use:
1. System Entities
As of now we support 4 types of system entities:
- Integer: Entity with data type natural numbers.
- Double: Entity with data type Double.
- Date: Detects date and returns result in epoch.
- Color: Detects the color. (Modifiable)
- Asterisk: When nothing else works. This acts as a wildcard and can match with anything.
More values can be added to Modifiable system entities. For eg. if AI Understander doesn't detect any certain color say Pitch Black, then brand can create an entity, with name My Color, extending Color system entity from platform and add Pitch Black as one of the value.
We are open to add more system entities. If you need us to support any system entity then please inform us.
2. User Defined
These are the entities defined by users. User can define entities from Train Tab.
Confidence
Confidence gives measure of how much AI Understander is confident about the detected Intent.
Valid range: 0 < Confidence <= 1
Chitchat
We also have support for chitchat. Chitchat understand basic user statements like Hi, Hello, Wassup, Bye, Thanks, Great Job, Who are you etc. Chitchat , is not provided by default in API , if you have a requirement for chitchat then please contact us.
API
As of now we are releasing private beta of Query API.
For beta users we won't be applying any rate limit restriction. Find the reference documentation here.