Handling Irrelevance with Teneo

Just like an employee who works for a business and provides a specific set of skills to support the overarching success of an organization, virtual agents are also incorporated in order to provide a specialized level of support.

While developing your virtual agent solution, you want to keep that in mind, enabling your bot in the area relevant to its role only. Moreover, you want to communicate to the users what to expect from the virtual agent, and you must prepare the bot with replies to out of domain (OOD) requests .

In this article, we explain how you can handle irrelevant topics using Teneo Platform. With some basic strategies such as: initially setting expectations and helping your bot to be prepared for gibberish input or irrelevant topics, and semi-relevant topics.

Set Expectations for Users of Virtual Agents

Setting expectations is always appreciated and there are different ways to do that. The most efficient way is found to be quick messages in the beginning of a conversation with a new user. Where the bot would display or recite the topics that are relevant and the tasks that it can perform. The topics can be provided as a simple list of suggestions that directs users to relevant topics.

Teneo Web Chat supports various rich message types that you can use to enrich your conversations with buttons, lists, cards, etc. You can add these messaging elements to outputs in Teneo Studio. They can be used to display suggestions of topics that the virtual assistant can help with. These elements can be styled to meet branding and user experience designs. The image below is an example of using Buttons in the greeting message which is the first message sent to the user.

The image below displays the way the output message is designed. Using Teneo Web Chat’s easy messaging formatting, buttons are added as Output Parameters with a title and different button options separated with a bar “|”.

Irrelevant User Input

Sometimes the user may start typing Gibberish or randomly ask irrelevant questions. Many irrelevant requests will be caught by flows included in the Teneo Dialogue Resources (TDR), Including a SafetyNet flow as part of these resources. TDRs can easily be imported when you create your solution. You can always modify and edit the flow to best fit in your solution project and your bot’s personality. You can read more about designing your bot’s personality in this article: When a Chatbot is Born.

The image below illustrates the way a bot responds to nonsense input.

safetynet for Gibberish and OOD

In the animation below, you can have a look inside the SafetyNet flow that is part of TDRs. Different messages can be stored as bot responses, providing a natural-like conversational experience.

SafetyNet Flow in Teneo

Semi-relevant User Input

Not all topics are completely OOD. Some topics are semi-relevant, and some input may be distantly related to a relevant topic.

For instance, a barista bot should catch input where the user mentions any sort of drinks and guide the user to the actual drinks that are served in that café. In the image below we see how the bot catches the input even though the user intent is not clear yet: is the user asking to order or for the price? However, the bot responds in a suitable way to bring the user back to tasks that are within the domain of the bot.

Area Specific SafetyNet

In Teneo, you handle this type of input by creating an Area Specific SafetyNet flow. Check out this practical example to do that and handle area specific input.

Virtual assistants are domain specific agents and their field of expertise needs to be clear to their potential users. With Teneo platform you have the necessary components to build solutions with clear messages about their domain. Moreover, you can gear them up with the needed flows to handle out of domain and semi-relevant topics.

1 Like