Connect with us

Hi, what are you looking for?

Apps

Due to Facebook’s new speech recognition technology, there will be no need for human transcription

Due to Facebook's new speech recognition technology, there will be no need for human transcription

At present, all large technology companies regard artificial intelligence (AI) as the future of technology and are working in this direction. The technology giant Facebook is also doing its best to develop artificial intelligence technology. The smart technology we use every day in smartphones, cars, speakers, etc. Speech recognition is still in progress, and many technology giants are improving it. Facebook marks a breakthrough in speech recognition and learning. The company has developed a speech recognition method that does not rely on text-to-speech for speech recognition.

In the past, people with traditional systems had to transcribe every record in every language, but this new speech recognition system no longer needs it. Facebook’s new unattended system can learn directly from the voice of human language and make the system better understand what a human conversation sounds like if it also saves a lot of time because it decrypts every message Recording takes several hours and working time. The generator generates a new data instance, and the discriminator evaluates its authenticity.

In Facebook’s new speech recognition model, the speech pattern output by the generator will be completely messed up when passing through the discriminator, and Facebook will insert other hand-written text to help the generator understand the difference between the computerized result and the actual result difference. Repeat this process. Until the output matches the actual text. The technology giant named its model “Wav2vec-U” (U stands for “unattended”) and began testing it.

The model was tested in Swahili, which is used in the Kyrgyz Republic of Central Asia. The test results show that compared with the next best-unsupervised method, the error caused by the system is reduced by 63%. The current unattended model is as accurate as of the controlled model a few years ago, which indicates that the model has been replaced. To speed up development, Facebook created a code for Wav2vec-U, which can be found on GitHub. A new era of speech recognition has arrived, and it will be good for Facebook because it will help democratize technology and achieve Facebook’s goal of connecting billions of people with their favorite language.

Due to Facebook's new speech recognition technology, there will be no need for human transcription

credit: Digital Information World

LAST THOUGHTS:

In the past, people with traditional systems had to transcribe every record in every language, but this new speech recognition system no longer needs it. In Facebook’s new speech recognition model, the speech pattern output by the generator will be completely messed up when passing through the discriminator, and Facebook will insert other hand-written text to help the generator understand the difference between the computerized result and the actual result difference. The technology giant named its model “Wav2vec-U” and began testing it.

Content Protection by DMCA.com
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

You May Also Like

News

Phishing is one of the most well-known types of cyber attacks that ordinary people may eventually suffer. This has a lot to do with...

Apps

Instagram appears to have implemented a new feature that allows users to verify their custom insight posts on the web or desktop version of...

Apps

The clubhouse is one of the fastest-growing applications in the world. Since its release, the app has received a lot of competition, whether it’s...

News

As we all know, the Google Play store has real malware problems. Although the most popular applications that can be purchased or downloaded on...