How AI is changing the way we learn languages

During the week of July 26-28, join executives. Top leaders talk about topics such as artificial intelligence, edge, and more. You can reserve your pass now.

One of two scenarios are usually what we think of when we think about voice recognition and artificial intelligence. If you buy the right kind of lawn mower, the first thing you're going to do is sit at home and listen to your conversations. The second scenario is a lot of auto-subtitling of our videos and TV shows.

There are some exciting developments happening in the voice recognition space. It is now possible to create programs that can analyze and score speech thanks to advances in artificial intelligence. They can do this on a number of criteria.

The transformational power of this ability to score speech is found in the language- learning and education spaces. Imagine a world in which a human doesn't have to teach. Imagine if that could be done in real-time. That kind of technological development would save huge amounts of money.

With the correct technology and models, any language student can theoretically receive feedback in real-time on how they are speaking, whether their English pronunciation is correct, and how to improve it. This is similar, but not the same as other artificial intelligence speech applications, such as automatic speech recognition, where an audio signal is received and a text is output.

An optimal system for this type of AI model requires the following five key components:

• Audio preprocessing that handles raw audio signals coming from different platforms
• An Artificial Neural Network (ANN) that receives a processed audio signal and produces embedded representations of the speech
• A post-processing layer that constructs human-readable evaluation
• An application-composer layer that maps the evaluation to product-feature needs 
• A proprietary system that monitors the quality and performance of the production system

In order for a system to give real-time feedback, the end-to-end time should be less than a second. This means that any core Artificial Neural Network only has a few milliseconds to respond, which presents a challenge in itself because this is a model with hundreds of millions of parameters.

One way to counter this is to use phonemes, which are units of sound in a language that distinguish one word from another. English has 44 phonemes, made up of 20 vowels and 24 consonants.

An artificial intelligence system can give feedback on how well a user's sounds are or how close they are to incorrect sounds. A system can give a score ranging from 0 to 100 on the phonemes. The platform can score the two words: /f/ and /l/ The full sentence could be scored by it. It is possible for imperfect pronunciation to match what it sounds like.

The systems are becoming more popular. Companies are able to use pre-trained models and invest a lot in the fine tuning of them. The key to model selection is the use of in-house knowledge about spoken English learning, engineering capabilities and deep knowledge of models' strengths and limitations. Artificial intelligence can be developed that gives users immediate feedback on how they speak English, even if they don't want to.

Off-the-shelf services on the GCP can help to reduce operational costs and ensure stability. These kinds of technologies allow for real-time feedback to be given to learners when they speak.

These kinds of technological developments could have a big impact on the education space. The lowering of costs is one of the main benefits of seamless artificial intelligence. English language skills are the main obstacle to landing a job with an international company, not geographical location. If software can help someone become proficient in English at a more reasonable rate than human-to-human tuition, then they will be able to find work in the global workforce. Speech-recognition and language learning could be the future of the international talent market. We need to build it.

Thy Ntrn is the chief technology officer.

The VentureBeat community welcomes you.

Data decision makers can share data related insights and innovation.

Join us at DataDecisionMakers to read about cutting-edge ideas and up-to-date information.

You could possibly contribute an article of your own.

Data decision makers have more to say.