|Microsoft Adds Conversational AI Agents|
|Written by Kay Ewbank|
|Thursday, 09 May 2019|
Microsoft has shown off new virtual agents that are to be made available to developers. The agents were demonstrated at this year's Build conference.
Intelligent agents that can handle human language in a realistic and useful fashion are a goal for several companies. In Microsoft's case, the demo showed its Cortana virtual assistant conversing in a more realistic fashion.
Microsoft says the demo moves beyond the restrictions of self-contained commands to a truly conversational experience.
The traditional approach to virtual agents relies on the system having a database of what the user might say and what the reaction should be. This makes it difficult to carry the context of one interaction into the next, or to have what appears to be a 'conversation' rather than issuing an instruction or asking a question and getting a response, then starting again from scratch.
Last year, Microsoft acquired a company called Semantic Machines that specialized in research into conversational AI. The technology from Semantic Machines has been used by Microsoft researchers, to put together new conversational AI technology that remembers what has been said in conversations to give a better experience. Microsoft says the technology
"crosses skill boundaries, connecting together back-end services, both within Microsoft and externally."
Unlike conventional approaches that combine a set of rules and code, the new technique is based on data and machine learning. Microsoft says the experience is:
"completely natural with multiturn dialog, spanning multiple domains and, importantly, working across multiple agents."
The natural language technology in today’s intelligent assistants such as Cortana makes use of machine learning to understand the intent of a user’s command. Once that intent is determined, a handwritten program – a skill – is triggered that follows a predetermined set of actions.
For example, instead of executing a hand-coded program to get the score of a football match, the Semantic Machines approach starts with people who show the system how to get sports scores across a range of example contexts so that the system can learn to fetch sports scores itself. There's a good description of the current research and how it works on the Microsoft AI blog
The technology is going to be integrated into a number of products including Cortana, and made available to developers in the Microsoft Bot Framework and Azure Bot Service.
Microsoft also announced improvements to the Cortana Skills Kit for Enterprise at Build. This is built on top of Azure Bot Service and makes use of Azure Active Directory to manage building, testing and deployment of custom skills. The kit now lets multiple developers in a team edit and manage the Cortana channel registration configuration for the bot. It is now also possible to restrict usage of skills to certain groups or departments within the organization to ensure information is available to appropriate employees only.
or email your comment to: firstname.lastname@example.org