The KONTAKT project set out to create a context‑sensitive, eye‑controlled communication system for people with severe communication impairments. The core technical achievement was the integration of a lightweight, offline N‑gram word‑prediction engine with a grammar filter that limits suggestions to syntactically plausible options. Speech recognition was handled by an offline module that delivered satisfactory accuracy for the target user group, while online prediction models were deemed too resource‑intensive for the required real‑time performance. The system also incorporated user‑specific knowledge bases and environmental context gathered from microphones and cameras, although camera‑based context was deprioritised in the final demonstrator due to user concerns. A virtual‑reality headset equipped with an eye‑tracker was used to showcase an augmented‑reality prototype, illustrating how future devices could deliver the same predictive assistance in a more immersive setting. Throughout development, the team conducted two major iterative design cycles, each informed by continuous user testing and feedback. Interviews with AAC users and caregivers produced a scenario map that distinguished private from public interactions, leading to the inclusion of informal elements such as emojis in the interface design. An ethics, legal, and social implications (ELSI) report was produced in collaboration with the project partner, addressing data protection, pseudonymisation, and the legal handling of personal audio and visual data.

The project was carried out from 1 May 2020 to 30 April 2023 under the auspices of the Fraunhofer Society for the Promotion of Applied Research (Fraunhofer‑Gesellschaft zur Förderung der angewandten Forschung e.V.) with grant number 16SV8472. The lead development partner was seracom GmbH, which took responsibility for the overall system implementation. Fraunhofer IAO contributed the scientific and methodological expertise, particularly in human‑centred design, machine‑learning integration, and the development of the prediction algorithms. RehaMedia Handelsgesellschaft GmbH acted as an associate partner, providing specialist knowledge in AAC and advising on user requirements and market deployment. The collaboration combined expertise in user interface design, artificial intelligence, and assistive technology to produce a system that is both technically robust and closely aligned with the needs of its intended users.