Audio-to-audio translation for status holders

Challenge: How to let non-Dutch speakers communicate in their own language?
Audio-to-audio translation for status holders, in real-time, powered by AI

Status holders are required to go through a participation and immigration process (PIP) at the municipality of Amsterdam. However, employees at the municipality (klantmanagers) and status holders often don’t speak the same language. Klantmanagers speak mostly Dutch, and status holders often Arabic or Tigrinya. Translators are not always available or viable. They are expensive, take time to organize, are not always available in the short term, and sometimes make errors. Meanwhile, most status holders speak languages that have an oral tradition; hence verbal communication is preferred. There are barriers to communication between klantmanagers and status holders.

Objective: Better communication would improve the experience throughout the process for both. To achieve this, Switch AI develops speech technology that does audio-to-audio translation, in real-time! A solution built using the power of state-of-the-art AI technology.

Solution: Switch will build a software service that leverages speech recognition, translation and personalized speech generation AI models. The service will aid klantmanagers and status holders to communicate in their own languages. This service can be integrated with various applications developed by Gemeente Amsterdam, such as the PIP app.

Empowering low-literate citizens with AI (Tolkie)

Challenge: Wildcard – Propose any technological solution for any marginalised community
Empowering low literate citizens of Amsterdam with the help of A.I.

Amsterdam has 33.3% more low-literate citizens than average for the Netherlands. Tolkie has developed a tool that helps low-literate users to overcome obstacles in a complex text. Organizations embed our tool into their website and by doing so they enhance the accessibility of their online platform. Our current solution is developed together with low-literate people. In an iterative process of acquiring user needs, designing and developing a solution, testing with the target audience, adjusting the design, etc. we developed the tool we have now. Our tool currently operates on a per-word basis, meaning that the target audience gets reading aid when they encounter difficult words. However, complexity of a text lies not only in difficult words, but also in complex sentence structures, the use of jargon and proverbial language, the structure of longer texts, etc. During this challenge we are going to extend our reading aid and we want to achieve that with the use of AI and other modern technologies.