Regular expression (regex) entities let you define logical patterns that you can use to match and extract information from an input. Regex entities are great for complex pattern matching against a user’s input, or if you need to allow for specific variations in how a user might format or enter their input in a conversation. For example, you could use a regex entity to identify items such as a tracking ID, a license number, a credit card number, or an IP address from a string the user enters into the bot.

For more information on using entities in Power Virtual Agents, refer to documentation here:  Use entities and slot filling.

Go ahead and try this feature at powerva.microsoft.com. Your feedback will help us continue to build on and improve the functionality. We want to hear from you!

Thank you,

Jeff

The post Regex (regular expression) entity support in Power Virtual Agents appeared first on Microsoft Copilot Blog.

I’m thrilled to announce the long-anticipated Topic overlap detection and Topic suggestions from chat transcripts AI features that are now available in public preview.

To use the features, simply go to Manage > AI capabilities, and turn these two features on.

Watch the below video to understand more on how to use the AI-assisted bot authoring features to optimize your chatbot.

Note: These features are currently in public preview, only available in English language PVA bots. 

The post AI-assisted authoring: detect topic overlaps and get topic suggestions from bot sessions appeared first on Microsoft Copilot Blog.

Keep reading to learn the details on how the newest release of AI capabilities will enable Power Virtual Agents to provide a more personalized and natural chatbot experience.

Topic overlap detection

With the amount of chatbot content growing, topic-overlap inevitably happens, resulting in the bot asking “Did you mean” clarifying questions more often.

Today, we are announcing an AI-powered topic-overlap detection tool which helps bot authors to identify and refine overlapping topics, reducing the need for the bot to ask clarifying questions.

Suggestions from chat transcripts to address gaps

Today we are also announcing a new topic suggestion feature that generates topic suggestions from chat transcripts. What this means is that as users chat with the bot, the system will analyze chat transcripts that don’t trigger a topic and suggest new topics that will help makers address those gaps. For example, when a lot of users are talking about “Thanksgiving hours,” the system will automatically suggest for this topic to be created.

Bots get better over time

With continuous learning in Power Virtual Agents, each conversation will make the next one better. Using signals from responses coming from the “Did you mean” questions, the bot will learn automatically, removing the need for the bot to have to clarify the same question over again. Thanks to this new AI capability, the more users interact with the bot, the better it will become when addressing questions.

Personalized conversations

Through the course of conversations with a user, the bot will reuse information from Microsoft Graph and Microsoft Azure Active Directory, leveraging it to enhance and personalize future conversations. For example, if a user mentions their name, email, or zip code, these properties are stored and leveraged in later conversations without having to re-prompt the user.

In the near future, Power Virtual Agents will also be enabled to add User Graph personalization. Data that is stored in Microsoft Dataverse or populated by other apps (for example, Portals and Microsoft Dynamics 365 Customer Service) will be able to be leveraged as context variables for the bot to use.

Cutting-edge NLU model boosts topic triggering performance

Traditionally, intent triggering is formalized as a multi-class classification problem, in which the model is highly associated with the known categories; any change to these categories will require building a new model.

Power Virtual Agents hosts multiple AI models and AI capabilities on a single service, the core of which is a transformer-based natural language understanding (NLU) model. Different from the traditional approach, Power Virtual Agents language understanding model uses the example-based approach powered by a deep neural model. Such large-scale model only needs to be trained once with massive amounts of data using AI supercomputing, and can be later used for specific tasks with just a few examples  and no further training. This is part of Microsoft’s AI at Scale initiative—which basically changes the way AI is developed. It provides an intuitive way for bot makers to work on the bot content simultaneously and confidently, without having to involve AI experts.

Delivering on our vision

Our mission is to democratize conversational AI, empower every individual and every organization on the planet to easily build, manage, and use intelligent bots—all without requiring any coding or AI expertise. To realize this vision, we will continue to introduce advanced AI capabilities in Power Virtual Agents, allowing bots to improve without any maker intervention.

• Dive deeper into the newest release of AI capabilities enabling Power Virtual Agents to provide a more personalized and natural chatbot experience.
• Begin your journey with Power Virtual Agents now by starting a trial today.
• Be sure to join our Power Virtual Agents community to share your ideas, provide your comments, and help drive the future direction of our product development.

The post Optimize your chatbots with new AI features for Power Virtual Agents appeared first on Microsoft Copilot Blog.

On some occasions, you may want to customize the bot behavior when no intent is triggered. For example like searching knowledge base articles or routing to existing bot services. This can be achieved by adding the System Fallback Topic from Settings. 

Once the System Fallback Topic is added, you can find it alongside with other system topics in your topic list. You can edit it just like you can edit other system topics — add or remove nodes, redirect to other topics, call services using Power Automate or Bot Framework Skills. There’s a {x} UnrecognizedTriggerPhrase variable in the topic which carries the last user utterance. It can be used as a context when calling other services.

You can find more information about System Fallback Topic here.

The post Customize the bot behavior when no intent is triggered appeared first on Microsoft Copilot Blog.

By specifying the type of entity to identify in a ‘Question’ node, the bot can extract and remember specific type of information and save it to a variable.  When having a conversation with end-users, the bot will actively listen to user responses, and always try to identify entities upfront to avoid asking unnecessary questions.

Check out this brief video to learn more about how to leverage smart entity extraction and proactive slot filling to build a more intelligent bot.

We’d love to hear about your experience with this feature! Please visit our community forum at https://aka.ms/PowerVirtualAgentsForum and share your feedback.

If you have any idea requests, please submit them at https://aka.ms/PowerVirtualAgentsIdeas.

Thanks!

The post Smart Entity Extraction and Proactive Slot Filling appeared first on Microsoft Copilot Blog.