NeuralSpace completes its $2.8m seed round and brings voice AI to local languages around the world
$2.8m seed round led by Merus Capital and GoHub Ventures, with participation from APX, Techstars, and others.
NeuralSpace develops voice and text AI for locally spoken languages in Asia and Africa, offering their AI models in more than 100 languages.
After seeing traction for their text AI models in the Middle East and southern Asia, the focus will be on developing voice AI for the next 12 months.
The $1.1 million add-on investment completes NeuralSpace’s seed round of $2.8 million seed round, which is led by Merus Capital and GoHub. Further investors are early-stage investor APX, Techstars, Verissimo and a few impactful angels.
The investment will help NeuralSpace to scale its voice AI technology alongside its existing services, which went live as a self-serve toolkit in January 2022 with Language Understanding in 90+ languages, and multiple auxiliary functionalities such as automatic data augmentation, automatic data set conversion (from Rasa, Google Dialogflow and Microsoft LUIS), language detection to capture the language preference of a user automatically, and translation to convert a user’s input whenever needed for additional context.
NeuralSpace now aims to develop the most accurate speech models (both Speech-to-Text and Text-to-Speech) for locally spoken languages. Furthermore, it develops pre-built end-to-end products such as video localization (or automatic overdubbing), which are combinations of the existing speech and text services that are already live on the NeuralSpace Platform as standalone features.
What is NeuralSpace?
NeuralSpace is a Natural Language Processing (NLP) company specializing in local, or low-resource languages. The NeuralSpace Platform is a collection of proprietary language models that can process up to 100 different languages. The latest funding will be used to double down on the voice AI development at NeuralSpace, which will include building models for mixing languages (such as Arabic-English, Chinese-English, Spanish-English or Hindi-English), and significantly increasing the accuracy of automatic speech recognition (ASR) models in locally spoken languages, compared to current market leaders.
Using the NeuralSpace Platform does not require any machine learning expertise, and all that is needed is a handful of data to train and continuously improve each user's custom models. The NeuralSpace Platform is a no-code, modular user interface and each of its services, from Natural Language Understanding (NLU) to Entity Recognition, Speech-to-Text, Machine Translation and Transliteration can be taken as a standalone product, even installed on-premise if required. In the end, users only need to connect their custom models with REST APIs.
NeuralSpace’s co-founder and CEO Felix Laumann points out that “training latest deep learning models accurately on very small data sets, which any local, low-resource language usually suffers from, is a very challenging task because models are usually designed to perform well in English with terabytes of available text and speech data but not in languages with a few hundred megabytes of data.”
Private Cloud Deployment
NeuralSpace offers to deploy the NeuralSpace Platform on-premise (or “private cloud”). It has already successfully done this with Hello Ebbot, a Swedish chatbot development company operating in Scandinavian and other European languages, where the NeuralSpace Platform is hosted on infrastructure arranged and provided by Hello Ebbot. It allows Hello Ebbot to comply with GDPR requirements as customer-owned data are not shared or processed outside of the private cloud that is owned by Hello Ebbot and hosted in the European Union.
One of NeuralSpace’s most important propositions for any developer is that all NLP requirements can be fulfilled within one single platform and switching between multiple vendors with different data formats for different NLP-related features is not required anymore. The NeuralSpace Platform provides various NLP-specific services like Language Understanding, Entity Recognition, Machine Transliteration, Transliteration, Speech-to-Text and many more in one place so that developers do not need to think about handling different services, payments, and contracts.
One of the key advantages of the NeuralSpace Platform is its scalability. It is designed in such a way that it can linearly scale to millions of requests every hour, which is achieved by a proprietary compute-load watcher, data allocator, and training scheduler. The scalability is extensively evaluated through load testing techniques by simulating a large number of requests the NeuralSpace Platform receives at the same time, providing statistics around the limits for a given number of model replicas, and providing an estimated cost for the infrastructure required for the given load quantities. Thus, it makes it easy for any developer to assess the economic feasibility of using NeuralSpace without having to go through a lengthy process of experimentation first.