How To Train AI Chatbots on FAQs

How To Train AI Chatbots on FAQs | Building Virtual Assistants on the Kore.ai XO Platform

You’ve seen them all over the Internet. We’re talking about FAQ’s, or frequently asked questions. You can even use your chatbot or Intelligent Virtual Assistant to answer customer questions. Chatbots for FAQ’s don’t just appear on a web page magically. They have to be created and trained to anticipate all of the questions that a web page user might have about an organization or topic.

Training Intelligent Virtual Assistants to Answer FAQs is Valuable

Training AI chatbots on FAQs for a website is valuable because it allows the chatbot to quickly and accurately answer common customer questions, reducing the workload on human support staff and improving customer satisfaction. By automating the handling of frequently asked questions, the chatbot can also provide 24/7 support, improve response times, and increase the efficiency of customer service operations.

You will do this by leveraging your Knowledge Graph Engine to build the FAQ Repository. You start with the Kore.ai Intelligent Virtual Assistant control panel.

The Kore.ai platform provides question-answer sets to relevant nodes in the hierarchy to deliver an intelligent FAQ experience to the users. You can also add synonyms and channel-specific responses.

(Source: https://kore.ai/platform/virtual-assistant/knowledge-graph/)

There are three main steps to training your FAQs

Step One — Build the ontology

Step Two — Add keywords

Step Three — Add synonyms

Advanced Topics

There are some advanced topics that we can cover but we’re not going to discuss them in this video. You can read more in our developer documentation. We will also have some more video content on advanced topics coming out soon.

Creating A Knowledge Graph

You can find the Knowledge Graph underneath conversational skills. Then you can jump into the actual FAQs that we have previously built out.

We can see that we have a couple of FAQ’s around a few different topics. In the chatbot controls you can see the ontology that we’re building out called ‘About Us’.

What is ONTOLOGY?

A set of concepts and categories in a subject area or domain that shows their properties and the relations between them.: “what’s new about our ontology is that it is created automatically from large datasets” “we’re using ontologies to capture and analyze some of the knowledge in our department”. — Google Dictionary

Another way to look at Ontology is that when we have common topics that fall underneath a certain umbrella, we can add all of those FAQ’s to that umbrella to provide a higher degree of accuracy when we’re receiving FAQ’s from users.

As you build out this hierarchical structure we see that we’re adding a little bit more hierarchy and some more organization, but then what we can do is set the ontology aspects.

Determine The Intents

Looking at all of the intents that are created and we see that with ‘Location and Hours’, those both would fall underneath an ‘About Us’ or ‘Company Info’ type of FAQ. On our ecommerce bot we can add a node and we can label it as ‘About Us’.

Creating A Node

We can create that node and then look at location and hours. We find that because those both fall within ‘About Us’, we add them there.

As we add hierarchy and a little bit more organization we are addressing the ontology aspect.

Adding Keywords

Next you can start adding some keywords. If you add the keyword ‘Hours’, which is pretty simple, you find that two in the ontology list are very similar — ‘What are the hours’ and ‘What are your hours?’.

So we’re going to add ‘Hours’ as the keyword for these two. Then we’re going to add ‘Store Hours’, or we can just do ‘Hours’ for all of these four FAQ’s, these four different ways of talking about the store’s hours.

That way you make sure that the FAQ’s are even more accurate based on if a keyword is identified within a user’s query.

Add Synonyms

The final step in the FAQ process is to add synonyms. A synonym is just a different way of saying the same thing by using a different word. So within the ‘About Us’ FAQ we can add ‘Schedule’ as a synonym to the ‘Hours’ FAQ. Time is another synonym that can be added.

If somebody asks what your schedule is, or what time do you open, then we know that they are also still talking about the store hours. This way we can surface the appropriate material. Click ‘Save’ and you are done adding synonyms.

Managing Context and Traits

There are some other things that you can do such as using traits and also managing context.

In one e-commerce example — a children’s clothing store — terms such as ‘Payments’, ‘Return and Exchange’, and others fall within the same node, so they can be entered and created for each path.

Some of these more advanced topics such as traits and contexts are going to be covered at a later time on the Kore.ai web site, or you can read about them in the developer documentation.

Final Step

The final step is to click the ‘Train’ button in the upper right hand corner of the developer dashboard, and that completes the process of training the knowledge graph and training FAQs.

This is a quick introduction on how to train intelligent virtual assistants on FAQs. You can read more about the Dialog Builder in the Kore.ai Developer Documentation and at https://kore.ai/platform/virtual-assistant/dialog-builder/.

About Kore.ai

Kore.ai is the leader in conversational AI platforms and solutions, helping enterprises automate front and back-office business interactions to deliver extraordinary experiences for their customers, agents, and employees on voice and digital channels.