Are you one of those people that likes to customize your pizza order, or ask for a decaf-double shot-skimmed milk-cappuccino? We at Chatlayer.ai have you covered!
Our launch of composite entities is ground-breaking news for chatbots! Why so? Thanks to this new feature, users can now present even more complex requests to bots and get the exact thing they asked for.
Take a look at how a simple request improved by using composite entities:
User request: “I want to order 1 pizza Pepperoni and 2 pizza Margherita”
- Chatbot answer without composite entities:
“Great, we will send you 1 pizza Pepperoni” or even worse
“We will send you 2 pizza Pepperoni and 1 pizza Margherita”
- Chatbot answer with composite entities:
“Great, we will put 1 pizza Pepperoni and 2 pizza Margherita right in the oven”
Thanks to composite entities, it is now possible for bots to understand complex expressions with multiple entities. Before, values included in a user request were floating around without being attached to their entities. This limitation often caused missing key parts of the user’s input.
How do composite entities work?
When users ask questions to a chatbot, the NLP algorithm starts extracting different values from each request, such as quantities and phone numbers. Chatbots can find an exact match for most of these items, but some entities need context to be framed.
For example, if you are not a real Italian, you might want to order “a pizza Hawaii”. The fact that “Hawaii” is preceded by “pizza” is crucial context that helps the bot understand that this is a pizza type and not an island in the Pacific Ocean.
These entities are directly extracted from users’ questions by the Chatlayer NLP engine. Are you familiar with entities? If not, take a look at our entities guide.
However, if multiple quantities and different pizzas are mentioned, it becomes very difficult to link the correct number of pizzas to the correct type of pizza. The chatbot doesn’t know which number is connected to which pizza type, so it will understand the wrong order.
The solution to this problem is composite entities. This new feature allows you to define a new group entity, which consists of multiple entities that follow a pattern.
For example, you can define a composite entity “Pizza Order” which always consists of a number, followed by a pizza type. This way, you will get two composite “Pizza Order” entities with the correct number and pizza type linked together. The result? Now you can order “1 pizza Pepperoni and 2 pizza Margherita” and the bot will get your order right!
How does this impact the user experience?
After all this technical talk, what’s in it for the user? Well, quite a lot!
In fact, until today chatbots were only able to understand one attribute per item included in each user request. They couldn’t understand customized orders and elaborate conversations.
Now it is possible for users to give chatbots more complex and personalized information. The result is a more human conversation and a better user experience.
Users can directly experience the difference in the algorithm, like in the following cases:
Example #1 – Virtual shopping assistant for e-commerce website
In this scenario it is now possible for customers to describe a specific outfit to the chatbot and receive a recommendation tailored to this request.
Example #2 – Food delivery app
When using chatbots for food delivery, users can now order multiple items with different extras in a single message. The chatbot will get their order straight away.
Example #3 – Public services portal
Composite entities are useful also in service applications, like a virtual police assistant. Here people can submit detailed information about different individuals and the chatbot will match the right attributes in the correct order.