Back to search

FORNY20-FORNY2020

STUD: DeepSign AS

Alternative title: DeepSign AS

Awarded: NOK 1.00 mill.

Google translate can translate between more than one hundred different spoken languages, but not for a single sign language. Where are all the sign language processors? Automatic translations between languages have made it easier than ever to communicate across cultures and to exchange information. Today you can easily make yourself understood in more than one hundred spoken languages with the help of translation tools such as Google translate. Nevertheless, a large group of languages have been left out of these technological benefits, the sign languages. In 2022, we can still not automate translation for a single sign language anywhere in the world. Sign language is gestural-visual language that uses facial expressions, body language, signs and sight to communicate. Automatic production of information in sign language must therefore be based on a visual representation - for example through human-like avatars on screens. Sign languages are full-fledged languages that have been developed by and used by people with hearing impairments but are also used by people with disabilities that make the acquisition of verbal language challenging. For the latter, the availability of sign language in public spaces and with people who use spoken language can be crucial with respect to ones quality of life and social opportunities. In a world full of screens, it is therefore not hard to imagine the enormous potential of automating sign language translations. Information in public spaces can be broadcast in sign language, so everyone receives the same messages. Conversations between a deaf employee in a store and a hearing customer can be exchanged seamlessly. Subtitles on TV can be offered in both sign language and written language, on all programs. In this way, everyone can be offered the same terms and we can break down prejudices across languages. So if this technology has such great potential, why do not we see sign languages everywhere? It is not easy to give a good answer to this question, but a possible reply must be based on both the technological, social and economic circumstances that today constitutes the framework for research, development and business. The world today is inclined to facilitate universal technologies that can be used by billions around the world. Capital tends to search for companies that offer simple solutions with high potential for automation - companies that find the ingeniously simple answer to a problem most of us did not even know we had. For example, the search engine methodology behind AirBnB or the database structures behind food delivery services. Sign language processors are far from a simple technology that solves a universal problem. At the core of the need for sign language processors are complex individual circumstances, a history marked by oppression and prejudice, as well as fundamental challenges at the forefront of research in machine learning, computer vision and animation technology. A person's needs for languages cannot be generalized. Innovation processes for technologies that primarily support minority needs are decelerated by a lack of capital and interest in a general market. So how do we resolve the conflict between capital and warm technology? At DeepSign we work for a warmer future for technology and our communities. Today's capital-driven innovation processes need a profound change of attitude. We fight for technology to be developed with a humancentric focus, and not processes that uses our humanity with the sole purpose of economic growth. Today is the time to ask ourselves whether we should let the forces of the market define the foundation of our common future. In 2021 and 2022 DeepSign deployed funds allocated by the Research Council of Norway to start building a sector for Norwegian sign language technology. This work has focused on development of basic algorithms for translation between Norwegian and Norwegian sign language and development of human-like avatars that can translate in real time. Part of the work has also focused on connecting Norway to the international research environment for sign language technology, government organizations and the deaf community of Norway, so that Norway can be among the first to automate translations for sign language.

DeepSign gikk gjennom høsten 2021 gjennom flere dyptgripende og uforutsette utfordringer, og bestemte seg følgelig for å avslutte prosjektet tidlig og legge ned selskapet uten å ha ferdigstilt sin teknologi. Prosjektdeltakerne har gjennom prosjektet mottatt uvurderlig erfaring innen produktutvikling, innovasjon, prosjektledelse, kundebehandling, merkevarebygging og vitenskapelig formidling. Vi håper at DeepSign har bidratt til å styrke bevisstheten i det norske markedet rundt teknologi for automatisk oversettelse av tegnspråk og behovet for denne typen teknologi. Vi holdt dialog blant annet med Norges døveforbund, språkrådet og andre tegnspråklige institusjoner om muligheter rundt teknologien, og med mange mulige kunder i både offentlig og privat sektor. Vi håper med dette at når et liknende prosjekt i Norge kommer på banen i senere tid, at prosjektet vil møtes med mindre motstand og en lavere terskel for dialog enn hva DeepSign selv opplevde under denne prosjektperioden.

Funding scheme:

FORNY20-FORNY2020