Beyond Coding: Watson Assistant Entities — Part 1, A Quick Overview
A Watson Assistant connects to users in many ways — via text on a screen, voice on a phone or even small, robotic bodies. But there are two common factors for each of these. The first is linking the assistant to the user interface. The second is the “brain”, the narrative, which drives the assistant.
The connections — such as Watson web-hosted chat widgets, webhooks, and channel connectors to platforms like Facebook, Slack and Intercom — allow the user to interact with our Watson Assistant. But that’s only half the story. The other half is the “brain” — the content and the intelligence — behind the interactive layer which interacts with customers.
By looking beyond the code, we can create positive, engaging and useful interactions.
Watson’s Brain
Watson Assistant has three components which work together to interact with users; the intents, the entities and the dialog. This is an over-simplification of the Watson Assistant system but it will suffice for this article, where we are examining Watson interactions, not Watson’s algorithms.
The intents allow a Watson Assistant to understand what the customer/end-user is asking.
The dialog component provides a Watson assistant with the narrative between itself and the customer/end-user.
In this article I’ll be looking at the third component, the entities. Entities allow a Watson Assistant to categorize and identify pieces of information to help make decisions.
But how can we use entities to ensure the systems we create are at their best? To help get us there we first need to understand what they are.
Explaining Entities
Entities. Are they synonyms? Are they patterns? The answer is Yes. Yes, they are synonyms. Yes, they identify recognizable patterns. But before we can explore how to implement entities, we should understand why they might be used.
An entity allows us to create categories of something. By something, I mean anything. An entity can represent several concepts, such as dates, locations, services provided, physical products, etc. Entities are only limited by the assistant’s domain, or topic, of discussion.
It’s important to understand an entity is used to group items or concepts to identify information.
But Watson Assistant has another use for entities — identifying recognizable patterns. This type of entity requires some knowledge of Regular Expressions, but this powerful feature allows you to extract information from a conversation.
For example, an entity can be used to recognize the pattern of an email address. The information can be identified and then stored for later use.
I’ll go into detail on each of these features of entities in later articles. For now we should understand that entities are a way of classifying a group and its members.
Entity Benefits
So why would we use entities if we already have intents to identify the customer’s/end-user’s question? Intents allow us to identify the customer’s query, but entities allow us to collect information, either for use with external systems, or to make decisions during interactions.
By identifying information, you can re-use the entity throughout your dialog structure. There is no need to create structures which identify each of the products we have for sale, or each of the end-user’s pets we want to learn about. We have a central location for these grouped concepts and they aren’t limited by the question being asked.
The fact that we can extract information from a question using multiplies the benefits. Understanding what the customer/end-user is talking about is vital, but being able to capture the information and personalize the conversation means you’re getting the most out of the assistant.
More About Entities
We’ve looked at a general overview of entities, but we haven’t scratched the surface. In future articles we will be looking at:
• The details of the different entity types
• How to use entities
• How to determine which entities we need
• Entity Best Practices
Read about the different types of entities in the second article of this series.
