Microsoft wants to enhance Teams with a name mispronunciation detector, according to patent

How often have you mispronounced a participant’s name in a Microsoft Teams meeting? Sometimes, we experience being in an online meeting where we try to pronounce a workmate’s name properly only to realize we might have gotten it wrong. It’s quite an embarrassing situation, to say the least.

But what if we told you Microsoft is taking the initiative to handle this challenge with an innovative feature: real-time name mispronunciation detection in Microsoft Teams?

The technology, described in a recent patent, automatically detects when you’re not pronouncing the name of one of your colleagues correctly and notifies you. Smart, right?

The feature is built to function smoothly in online meetings. It listens carefully, finding the names of people participating in what’s being said and checking if they match a reference. If a difference is noticed, it tells the speaker about the name mispronunciation during the Microsoft Teams meeting. But how does it do this?

The system starts working before the meeting begins. It collects the names and user identifiers of every person participating in this meeting. Then, it creates a special language model for the meeting to ensure it can correctly recognize and write down names when spoken aloud.

When a name is said out loud, the system starts its action: it removes an audio part related to that name and compares this sound to reference pronunciation, which matches the user identifier linked to the participant’s name. The comparison is made by complicated algorithms that evaluate how similar the name said out loud is to the reference, producing a pronunciation score.

The speaker is informed if the score goes beneath a particular line, showing a wrong pronunciation. This is not where it ends; the system provides an experience of practice that lets speakers repeat saying the name and get feedback until they say it correctly. This element is very useful at the time of communication and also for later use in learning how to say things correctly.

This method of providing real-time feedback for name mispronunciation in Microsoft Teams is both a technical achievement and a game changer. It requires intricate steps such as making custom language models, getting and contrasting reference pronunciations, and presenting instant notifications while maintaining privacy and safety considerations.

The feature’s capacity to adjust for human speech subtleties and variations in names from different cultures demonstrates the progress made in speech recognition and machine learning technologies.

According to the patent, the images in the paper use Microsoft Teams as an example of a platform that can benefit from this capability. Teams will most likely get it in the following months or years.

Microsoft’s real-time name mispronunciation detection feature is a modern, unique method for promoting comprehension and connection in our digital dealings. It’s good to see that technology when used correctly, can unite us more closely, making our actions appear more human rather than less.

Not only that, but this unseen assistant who guarantees that everyone’s name is pronounced correctly will spare us all from a great deal of embarrassment.

This is not the first time the Redmond-based tech giant has released patents for Microsoft Teams: in 2023, the company also revealed a patent that described a technology meant to personalize the Teams experience for each user.