Edge Natural Language Understanding

Try the technologyContact us

Getting familiar with the technology

Natural Language Understanding is a technology designed to extract the intent from human language, as a human would be able to basically. It gives insights to words a computing system is processing.

Based on neural networks, NLU is defining rules to associate language with meaning. Instead of using streamlined expressions in common speech recognition, NLU is freeing the way users can express their intents.


How to use Natural Language Understanding (NLU)?

Identify all the actions...

Decypher user intents from their voice with advanced language understanding. Offer freedom of use through flexible and large vocabulary voice commands, still embedded.

From the ASR: “Turn on the air conditionning in the meeting room.”
With the NLU, you could say: “It is too hot in the meeting room” and it would translate: “$intent: Turn on the $equipment: air conditionning in the $room: meeting room.”

...To improve your voice system

Scale this process to a complete voice-activated system to provide users with flexible voice commands, freeing their experience from streamlined commands that leads to instant frustration.

Indeed, most voice user experience are sacrificed because of a poor UI that is not teaching the user what needs to be said as a command, resulting in a system error.


Synergies with other technologies

Speak freely

Natural Language Understanding only processes text data. In a voice-oriented use case, and in order to leverage the intent recognition capabilities from the NLU, voice needs to be transfered into text.

A phonetic model coupled with a language model are transcribing the user’s voice into the most accurate textual transcription possible. This result is then processed by the NLU to detect the intents and the entities that are involved to perform the according action(s).

More natural voice commands

Rule-based ASR, through its streamlined architecture, is much more accurate in specific situations that can be expected. Unlike Free Speech ASR, it is extremely lacking language flexibility. In other words, users HAVE to say exactly the voice commands that have been set.

Rule-based ASR are usually based on grammars that contain tags (the equivalent of slots, entities and intents) that are inlined within the voice commands as placeholders. Natural Language Understanding in this situation is automatically processing intents from the given text by the ASR, which is greatly optimising the tagging process.


Why should you use our
Natural Language Understanding?


Our NLU models solely rely on the device capabilities to work, no Cloud/Server requirements.

Data Privacy

Since our solution does not communicate with any third party, data stay in the device.

No latency

On-device processing is the best option to avoid service latency that highly impact the user experience.


NLU is made simple with its VDK Studio interface that enables any developer into a natural language expert.


Multiple NLU models can be used as layers
to adapt to any use case. Customisation is centric.

Clear Pricing

Unlike its alternative Cloud solutions, our NLU models don’t bill each time they are processing data.

Adopting voice solutions in your business starts here

Get in touch with our team to shift your company in the Voice First world! Or try it now!

Companies that have chosen our solution

Our customers' feedbacks regarding our solutions

“Vivoka’s technologies have been integrated into our solutions dedicated to Human Factors in industry, guaranteeing performance and data confidentiality.”

Guillaume Jégou

Head of the Human Factors Technology Laboratory

"Vivoka’s solutions can run offline, on-device, to operate anywhere, anytime and with a broad language support capability which is for our products a distinct competitive advantage."

Paul Travers

Chief Executive Officer (CEO)

zorabots logo

"We partnered with Vivoka on several innovative projects on embedded Linux system, and delivered, in a very short time, multi-lingual natural voice interactions fully running “at the edge” to our customers."

Cédric Vincent

VP of Technology

For developers, by developers

Try our voice solutions now


Sign up first on the Console

Before integrating with VDK, test our online playground: Vivoka Console.


Develop and test your use cases

Design, create and try all of your features.


Submit your project

Share your project and talk about it with our expert for real integration.

Requirements & Quick-Start

How to develop with our NLU engine?

– On the Device

Fully-embedded voice technology for brands seeking the convenience of a voice user interface without the privacy or connectivity concerns of the internet. Full access to custom commands and the ability to instantly update command codes during development make voice-enabling your product fast and easy.

– On Premise

Get the power of cloud connectivity combined with the reliability of embedded voice technology. On premise (or hybrid) solutions ensure that your device is always-on and responsive to commands. Seamlessly push product updates and deliver a broader voice experience with the level of cloud-connectivity that best matches your product and users.

Our NLU feature includes 3 functionalities:

  • Intent extraction: Ability to extract the intention of a sentence
  • Entities extraction: Ability to extract zero or more entites from a sentence
  • Specialized to a custom domain: Fine-tunned on customer domain

The engine currently supports 5 languages: 

  • fra-FR → French of France 
  • eng-US → Enlish of United States 
  • ita-IT → Italian from Italy 
  • spa-SP → Spanish from Spain 
  • deu-DE → German from Germany

– Model & Resource Size: 30 – 50MB

– SDK Code Size: 10 to 65MB

– Supported Hardware: Microprocessor Units

– Supported Platforms:

  • Windows – x86_64
  • Linux – x86_64 | armv7hf | armv8
  • Android 6.0 (API 23)

Complementary Technologies

Discover other technologies in our stack

Speech Recognition

Turn human speech into text data that can be processed by complex systems.

Voice Biometrics

Seamlessly identify or authenticate users by recognizing their voice pattern.


Automatically generate life-like voices from text data in multiple languages and aspects.

Wake Word

Trigger speech recognition process by detecting a unique word or sentence.

It's always the right time to learn more about voice technologies and their applications