At Voysis we believe that we have suffered the constraints of traditional interfaces for too long. Recent advances in deep learning and AI mean that the vision of conversing with our devices and services naturally through speech is now possible.
This places us at the cusp of a fundamental change in human-computer interaction. Only when conversational enabling technologies are widely available and have human levels of accuracy will the paradigm shift. Ubiquitous availability of conversation capable devices and services will bring with it a host of new modalities – at Voysis we are excited to help make this a reality.
The state of the art technologies needed to realise this are typically in the hands of a few large companies that do not allow much, if any, customisation beyond their own use cases and services. This greatly reduces the accuracy, flexibility and mass adoption of voice based conversational interfaces. Voysis levels the playing field and brings this technology to the masses, empowering any company to integrate powerful voice conversational capabilities into their services and applications.
Most speech recognition and conversational interfaces are generic and attempt to understand what the user says without any deep domain specific knowledge that facilitates an understanding of the user’s true intent. The precise words spoken by the user are far less important than what the user actually wants. The problem is that self serve conversational platforms allow for only very shallow levels of integration or customisation, which significantly hampers the potential of such interfaces. Deep integration is the key to realising the promise of true conversational interfaces.
Modern speech recognition and conversational systems can deliver astounding performance across a specific set of domains. However, the lack of customisation available to third parties using these systems means that it is not possible to achieve the same levels of performance for proprietary domains.
Voysis is markedly different and free of such limitations. Voysis is the only full-stack conversational platform for creating a custom voice based natural language user experience. The platform delivers unprecedented levels of accuracy and performance through the incorporation of domain specific vocabulary and large data sets.
Using the Voysis platform, companies can have their own deep domain natural language voice interface which is unique to their brand by being generated specifically for their application, service, use cases and data.
The Voysis deep learning engine creates tightly coupled models across the entire stack. These models include Natural Language Understanding (NLU) and Automatic Speech Recognition (ASR), all generated using the highly performant Voysis deep learning engine which has been designed specifically for use in conversational systems. Unlike other approaches, which simply chain multiple models together, Voysis’ models are generated in tandem and work in unity, not as separate isolated components. This is the key to achieving beyond human levels of accuracy and is similar to how we as humans naturally understand language.
The models are generated based on the vocabulary and the large data sets that you provide (e.g. your full product catalogue, service directory, etc.). This game-changing approach means your Voysis service understands your customers true intent, not just the words they say, because it deeply understands the intricacies and vocabulary of your service.
This level of deep integration also allows the Voysis platform to generate customised services across a number of market verticals. For example, our eCommerce focused natural language search and discovery service which allows users to find products simply by describing what they are looking for, but that is a topic for another post.
Over the weeks ahead, we’ll be posting specific examples of domain specific voice experiences. Subscribe to our blog below to stay in the know.