Introducing a Paradigm Shift in Speech Technology.

Use Vestec’s speech recognition and Natural Language Understanding products to create a sophisticated voice-activated user interface for your product.

We support both keywords and natural speech interaction and also offer custom solutions in embedded, server or cloud configurations for a wide variety of applications in all major languages.

We feature the VASRE (Vestec Automatic Speech Recognition Engine) line of standards-compliant speech recognition engines and the EVA-SDK (Embedded Vestec Architecture – Software Development Kit) product for creating embedded speech grammars and speech recognition engines for various microprocessor platforms, As well, we have Natural Language Understanding products in the beta development stage

Vestec’s products represent a true paradigm shift, from the traditional pattern matching methodology to AI-driven quality and performance.

Driven by Natural Language Understanding, they feature keyword spotting combined with semantic intent and contextual analysis to accurately identify what a speaker is saying. This provides a number of significant advantages:

  • Menu navigation that routes the user to the correct node in the menu tree, thus eliminating the need to navigate through a series of intervening nodes
  • Interpretation of free-form text to populate content into any structured electronic record
  • The ability to intelligently detect, differentiate, and flag compliance irregularities, and
  • Transcription of spoken content for indexing analysis using domain identification

Our products can be utilized to create sophisticated speech-based user interfaces for any business process, in any industry, in any language, and on any platform: see our extensive list of applications. We support both keywords and natural speech interaction, and also offer custom solutions in embedded, server, or cloud configurations for a wide variety of applications in all major languages.

EVA-SDK Embedded Speech Recognition

Vestec’s EVA-SDK (Embedded Vestec Architecture – Software Development Kit) presents a cost-effective, small footprint solution for embedding highly responsive and accurate speech recognition into many applications.

The EVA-SDK is a set of software development tools that enables the integration of customized, embedded speech recognition capabilities with a wide variety of applications on the microprocessor unit platform of your choice; developers do not need to acquire specialized speech recognition skills. The possibilities cover products over multiple sectors, including consumer, industrial, commercial, medical, and more – for use anywhere and anytime without requiring an Internet connection.

For more information, download our information sheets for the EVA-SDK product and its integration with STMicroelectronics’ STM32F7® Series MCUs.

VASRE Server-based Speech Recognition

The server-based VASRE (Vestec Automatic Speech Recognition Engine) line of standards-compliant small, medium, and large vocabulary ASR Engines generate high recognition accuracy and are designed for a variety of different tasks and various platforms. Because they are standards-compliant, they are easily integrated into your existing product or service. We support multiple languages, covering country localization needs with exceptional handling of a variety of native and non-native accents.

For more information, download our Automatic Speech Recognition information sheet.

Natural Language Understanding (NLU) Engine

With Vestec’s NLU Engine, you can enable conversational interaction in virtually any business process, in any industry, in any language, on any platform including sophisticated functions such as menu navigation and form filling.

One of Vestec’s core capabilities is applying the latest in Artificial Intelligence (AI) research to our work in natural language understanding. Compared to traditional approaches, our AI approach increases semantic understanding by 10-20% while decreasing grammar development costs by 30-50%. The NLU Engine improves semantic interpretation through the application of sophisticated context-based fuzzy inference rules. It recognizes new contextual associations between words encountered and retains them to apply them intelligently in future situations.

You can create a Vestec NLU system by coupling a VASRE or any ASR Engine to a Vestec NLU Engine; Vestec’s NLU Engine is ASR, language, domain, and platform independent. To complement, use Vestec’s TTS Engine to convert text inputs into natural sounding spoken speech using efficient and robust synthesis technology.

For more information, download our Conversational Speech Technology information sheet.

Speech and NLU Toolkits – Eclipse Plug-ins

Vestec’s Knowledge Engine (KE) and Data Engine (DE) Toolkits are plug-ins designed to help developers create natural speech menu navigation and form filling solutions. They increase routing accuracy by 10-20% and reduce grammar development time and costs by 30-50%.

Backed by Vestec’s NLU Engine, the KE and DE Toolkits provide three unique features. Firstly, keyword spotting plus semantic intent and contextual analysis accurately identify what the caller is saying.

Next, menu navigation and form-filling technologies direct the caller to the correct node in the menu tree, and fill out all slots in one shot; this eliminates the need to navigate through a series of intervening nodes.

Then, the ability to create or change speech grammars in-house quickly and easily eliminates the expense and inconvenience of outside consultants.

Speech and NLU Toolkit – Web Application

In addition to the Knowledge Engine and Data Engine plug-ins, Vestec offers the web-based SLIC (Speech Lab In the Cloud) application for creating NLU models to enable conversational interactions with your connected electronics and applications.

Use SLIC to create a concept-based NLU model to identify keyword instances in the user’s natural language speech. Voluminous amounts of training data; instead, SLIC maps your list of keywords to specified nodes in a hierarchical navigation menu to allow direct access routing for menu navigation or form data capture.

With SLIC’s NLU modeling capabilities, your users speak in a natural conversational manner to your product, be it an electronics device or application. Your SLIC model understands the user’s intent based on the content and context, thus allowing your product to respond quickly. When necessary, SLIC even enables your application to prompt the user for more information.

For more information, download our SLIC information sheet.

Text to Speech

In Vestec’s MRCP (v1 and v2) and SSML standards-compliant Text-To-Speech (TTS) Engine uses a unique text synthesis methodology to generate natural sounding speech like never before. What’s more, you can configure speech rate and pitch level to allow for a more animated, natural sounding voice. Vestec’s TTS Engine can be implemented in accordance with your specific requirements, whether for embedded applications, desktop or network server applications. Our versatile C/C++ language API can be complied for whatever operating environment you are using.

Vestec’s TTS Engine includes a comprehensive dictionary for text normalization, including exceptional processing of abbreviations, homonyms, and numbers in the text input. We are also actively working on offering the capability to integrate or build custom dictionaries for your applications.

Vestec’s text synthesis technology achieves a higher degree of naturalness in speech than is possible with the shorter diphones used in the more common unit selection methodology. While most systems allow voices to be optimized for naturalness for speech rate, pitch, and cadence, Vestec’s TTS Engine is unique in that it allows the speech rate and the pitch to be adjusted independently, thus allowing for a more animated, natural sounding voice. Another advantage is a smaller footprint, which makes it easy for this technology to be deployed in memory-constrained embedded platforms.