The U.S. Patent and Trademark Office (USPTO) has just granted a new patent to Microsoft, indicating that it is developing an agency technology that reflects the user’s conversational style and facial expressions. The application, titled Linguistic Style Matching Agent, was approved by the USPTO on September 3, 2020.
(From: Microsoft / USPTO)
The patent’s inventors have written the names of many people, including Daniel J McDuff, Kael R. Rowan, Mary P Czerwinski, Deepali Aneja, and Rens Hoegen.
With the development of speech recognition and conversation generation models, chatbot/virtual proxy dialogue applications have become more and more popular.
At present, most of these natural language interactions still remain in the monotonous and constrained stage, the expression seems not natural enough, usually only suitable for transactional tasks.
But Microsoft’s new technology is designed to bring an end-to-end agent that can engage in more natural/multi-round conversations.
By recognizing and matching the characteristics of the user’s speech rhythm, the virtual agent can synthesize a more natural and matching language style. As a result, users can experience a real-life conversation style (voice) and even facial expressions (video).
When implemented, virtual agents can respond using a subscription script or turn to machine learning techniques to generate an open response mix. It is believed that as the technology matures, it is expected to lead to an overall improvement in human-computer interaction.