NeuralSpace, a London-based Natural Language Processing (NLP) company specialising in local or low-resource languages, has secured $2.8M in a seed funding round led by Merus Capital and GoHub Ventures, with participation from APX, Techstars (who also invested in Roleshare, Oper Credits and Simplifyber) and others.
How Will The Funding Be Used?
NeuralSpace launched voice AI technology alongside other services earlier this year thanks to recent funding.
This toolkit contains multiple auxiliary functions such as language detection, data set conversion, and data augmentation. It also offers Language Understanding in over 90 languages through a self-serve toolkit. Additionally, it allows users to convert their input into translated text whenever more context is needed.
How Was NeuralSpace Born?
While attending TU Kaiserslautern in Germany for their Master’s degrees, Ayushman Dash and Kumar Shridhar started a student project. In 2017, this project grew into NeuralSpace when Kumar met Felix Laumann the following year in Copenhagen while working at BotSupply. At that time, BotSupply was a recently founded startup.
Shortly after meeting for the first time in Berlin, NeuralSpace was created roughly a few months later in Kaiserslautern. Their first paying customer arrived in November 2019.
Accurate Speech Models
NeuralSpace aims to create the most accurate speech models for every language: Speech-to-Text and Text-to-Speech. Creating these models allows Google Home, Siri and Alexa — smart assistants — to understand more languages.
The NeuralSpace Platform already offers separate speech and text services. By combining these services, new end-finished products can be created— such as video localization, also known as automatic overdubbing.
The SaaS platform provides developers with APIs for NLP work that can be accessed without ackknowledge of machine learning or data science.
Speech-to-text, natural language understanding, machine translation and transliteration are all standalone products that can be purchased from the no-code interface. It can also be installed on-site if desired.
Five applications make up the NeuralSpace Platform, including Language Detection, Neural Machine Translation, NeuralAug, Transliteration and NeuraLingo.
NeuralSpace provides language support for 80+ languages spoken in India, Southeast Asia, Africa, the Middle East, Scandinavia and Eastern Europe. These are referred to as low-resource languages.
NeuralSpace co-founder and CEO Felix Laumann says, “It’s very difficult to train the latest deep learning models on data sets that are small and local. Any language that has a small data set usually suffers from this problem. This is because models are typically designed to perform well with terabytes of English language text and speech; they aren’t optimized to work with a few hundred megabytes of data for any language.”
