Easy FAQ Building
In a Conversational AI project, you will want to include many types of different flows, for example:
- Welcome Message
- Business Processes
- Disambiguation
- Out of Scope
- SafetyNet
- FAQs
In this article, we are going to take a look at the latter, FAQs. An FAQ flow typically contains a simple question-answer pair.
There is normally also no backend integration involved, nor deeper identification of user data. The difficulty lies more in the high quantity of flows which need to be created and how they fit in the best way into your intent recognition setup.
In the following, we will go through different options to handle your FAQs in Teneo, which can also be combined if that suits your project best. Let’s take a look.
FAQs via Teneo Flow Building
Building an FAQ flow in Teneo is very easy. You go inside your Teneo solution, select the folder where you want to create your flow and click on the Flow icon in the top ribbon bar.
This creates by default the content we need for an FAQ flow: one Trigger + one Output Node
Now you just need to add some example questions, define your Match Requirements for the Trigger, and the answer text which will be returned to the user. The TODO labels guide you here to fill out the required information. It is also best practice to give both the trigger and the output node a name, so that the flow design speaks for itself when you or a colleague come back to it.
Please follow the steps in this little tutorial for all details on how to set the flow up from start to end, you will see it’s a very fast and easy process.
FAQs via Bulk Import
Teneo’s Bulk Import functionality makes it possible to create FAQ flows outside of Teneo, in a simple CSV file. You just have to follow the expected import format to set up all the details around the Trigger and the Answer. Teneo will then do its magic and create the desired FAQ flows for you.
You can find a detailed explanation and a sample file with the expected format here: Use Bulk Import | Teneo Developers
The required CSV will look similar to this example:
and the Bulk Import can be accessed like this:
When you use the Bulk Import feature, a high number of FAQ flows can be created in next to no time.
FAQs via Question Answering
Microsoft’s QnA Maker is currently transitioning into Question Answering, so I will use the latter as name in the following.
Question Answering is available as service within Microsoft’s Language Studio and makes it possible to automatically create FAQ-style Knowledge Bases (KBs) based on structured information in documents. A wide range of formats is supported, such as pdf, docx, xlsx, txt and tsv. Crawling the document from a URL is also possible. Here you can find a Quickstart guide.
Let’s say we would like to build our bot’s FAQs around Elvis Presley, then we could select the following url for the source information: Elvis Presley FAQ | Graceland
and Question Answering would parse this structured information into the following question-answer pairs:
The service uses in this the identified question utterance to train its Machine Learning model for the corresponding intent. You can add yourself additional alternate phrases to further improve the model’s accuracy on a specific intent.
Make also sure to download our Connector solution from teneo.ai here. Some details on the latest Connector update are also available in a Forum post. Once the KB is created, it can be called as a service from the Teneo solution. Please note that additional costs apply here based on Microsoft’s pricing for the Question Answering calls, and the limits of the service have to be taken into account. Regarding your project’s Intent Recognition setup, Question Answering KB is running its own Machine Learning model and returns you its answer together with a confidence score. The latter can then be used inside Teneo to decide if the answer from the Question Answering service is a good candidate. You can find an example flow in our Connector solution which would be triggered on all Elvis Presley questions, since the connected Question Answering KB is dedicated to this topic:
This is just an example though, as always, you have all freedom to set up the Intent Recognition in Teneo to what fits best your projects needs .
Recommendations
As we have seen, there are many options in Teneo to easily create FAQ flows. Now the question is which one is the best option. The answer to this depends highly on what you want to achieve and what criteria are most important for you. So, here’s a quick overview for you to take the decision which best fits your project:
The good news is also, you can combine all three seen approaches in your project. Which option will you choose?