Person using GazeTalk eye-tracking system at a desk

Language rehabilitation & communication,
guided by your personal style

GazeTalk.AI enables people with aphasia communicate through multimodal interaction and practice evidence-based language therapy at home supported by an AI trained in rehabilitation. GazeTalk.AI is under development at the Technical University of Denmark (DTU), Department of Management.

DTU logo
DTU Management Technical University of Denmark · Copenhagen
What is GazeTalk.AI

A new kind of platform — built around the patient

Co-creation session with a person with aphasia
Co-creation of a personalized conversational agent with a person with aphasia

Aphasia is an acquired language disorder — most commonly caused by stroke — that affects the ability of speaking, reading, writing, and comprehension. Despite the importance of long-term and post-acute rehabilitation processes, the vast majority of people with aphasia often struggle with limited therapist availability, lack of personalization, and poor long-term adherence. Moreover, quality of life of aphasia patients and their family & friends often deteriorates due to breakdown in communication.

GazeTalk.AI combines an Augmentative and Alternative Communication (AAC) system with an AI trained in delivering aphasia rehabilitation in a platform utilizing multimodal interaction (eye gaze, head tracking, speech, and keyboard stroke) — helping patients to communicate via different modes, practice independently, and follow a personalised rehabilitation plan — while keeping clinicians informed through a real-time dashboard.

01 — AAC
Multimodal communication
Patients select words and icons through multimodal interaction — including eye gaze, head movement, speech, and finger input — accessible even for people with severe motor impairments alongside aphasia.
02 — AI ASSISTED Therapy
Adaptive conversation partner
An LLM-powered agent applies scaffolding, supported conversation, and repair strategies — calibrated to the patient's aphasia profile and session history.
03 — Human Digital Twin
Your own language replica
A personalised language model built from your conversation history captures your vocabulary, patterns, and pace. It tracks recovery over time, monitors fatigue through typing dynamics and gaze behaviour.
Our History

Two decades in the
making — and evolving

Original GazeTalk interface
The original GazeTalk — gaze-based AAC for ALS, early 2000s
GazeTalk AAC on iPad
GazeTalk AAC interface on iPad
GazeTalk.AI interface 2026
GazeTalk.AI patient interface — topic selection and exercises (2026)

GazeTalk was first developed at the IT University of Copenhagen in the early 2000s by Prof. John Paulin Hansen and colleagues — originally as a gaze-based communication tool for people with Amyotrophic Lateral Sclerosis (ALS) who had lost the ability to speak and move. The system let users compose text and communicate through eye movements alone, pioneered by foundational work on predictive gaze-based keyboards (Johansen, Hansen & Itoh, 2006).

Today, GazeTalk has evolved. Communication remains at its core — but the platform now adds a new dimension: language rehabilitation. Powered by AI, GazeTalk.AI supports people with aphasia through personalized & independent therapy, helping them not just to communicate, but to recover. This evolution builds on years of peer-reviewed research, including studies on mental fatigue measurement using eye metrics (Bafna & Hansen, 2021) and mental fatigue prediction during eye-typing (Bafna, Bækgaard & Hansen, 2021).

References Bafna, T., & Hansen, J. P. (2021). Mental fatigue measurement using eye metrics: A systematic literature review. Psychophysiology, 58(6), e13828.
Bafna, T., Bækgaard, P., & Hansen, J. P. (2021). Mental fatigue prediction during eye-typing. PLOS ONE, 16(2), e0246739.
Johansen, A. S., Hansen, J. P., & Itoh, K. (2006). Predictive, restricted on-screen keyboard for gaze communication. In International Conference on Computers Helping People with Special Needs (pp. 555–562). Springer.

Who it's for

Designed for everyone in the care journey

GazeTalk.AI serves the clinicians guiding recovery, the patients living it, and the caregivers supporting them.

For Patients
  • Communicate through multimodal interaction — eye gaze, head movement, speech, and finger input
  • Practice various therapy techniques at home, at any time
  • Personalised to your interests, vocabulary, and pace of recovery
  • Your therapist stays informed — remotely, in real time
For Clinicians
  • Real-time language metrics per session (words per minute, type-token ratio, error taxonomy)
  • Longitudinal trend charts and outcome tracking across weeks
  • AI-generated session plans — fully editable and overridable
  • Evidence-based intervention matching to the patient's profile
For Caregivers
  • Stay informed without being in the room — session activity and progress are always visible
  • Help the people in your care communicate more effectively through multimodal interaction
  • Reduce the burden of daily practice — GazeTalk.AI enables independent sessions at home
Immediate Aims

What we are working
to achieve now

In 2025, GazeTalk.AI received the Distinguished Innovator Grant by Novo Nordisk Fonden. Currently, GazeTalk.AI is in active pilot development. These goals define our objective to conduct a clinical feasibility study in the period of 2026–2027 in Denmark.

01Clinical feasibility study in DenmarkStructured pilot with Danish speech-language therapists and patients — collecting real-world data and clinician feedback.
02Evidence-grounded intervention matchingSelecting the right therapy technique for each patient based on their measured language profile.
03Local LLM for clinical environmentsPrivacy-compliant, locally-hosted open-source LLMs — no cloud dependency, hospital-network ready.
04Synthetic patient simulatorA validated model of aphasia patient language enabling safe offline testing of therapy algorithms.
05Patient engagement and gamificationAdaptive difficulty and motivational mechanics that sustain weeks of home-based practice.
06Longitudinal outcome measurementPredictive trajectory models that estimate patient progress and surface early signals of treatment response.
The Team

The people behind GazeTalk.AI

An interdisciplinary team combining expertise in human-computer interaction, clinical natural language processing, machine learning, and aphasia rehabilitation.

J
Prof. John Paulin Hansen
Professor · Principal Investigator
E
Ekky Tammarar Alfian, Ph.D.
Postdoctoral Researcher · Product Manager
P
Pouya Nikoui, MSc
Technical Specialist
H
Assoc. Prof. Haakon Lund
External Advisor
A
Anders Sewerin Johansen, Ph.D.
External Advisor
Open Positions/Projects

Help build the future of aphasia rehabilitation

We are seeking students and researchers from diverse backgrounds to collaborate with us on GazeTalk.AI. In particular, we actively seek DTU students to do a project/thesis with us. See opportunities below for further detail. Do not see a project that fits? We are always open to new ideas and collaborations. Reach out and let us explore ideas together.

Synthetic Patient Simulator for Stroke Aphasia Rehabilitation Exercises
OpenMSc / BScNLPML
Developing a Local-Based LLM for Stroke Aphasia Rehabilitation
OpenMSc / BScNLPFine-tuning
Gamification of Stroke Aphasia Rehabilitation
OpenMSc / BScHCIUX
Realizing the Human Digital Twin for Stroke Aphasia Rehabilitation
OpenMScMLNLP
UX Design and Co-Creation Workshops for Future GazeTalk Development
OpenMSc / BScHCIFigma
Modelling Fatigue in GazeTalk Using Text Input, Subjective Ratings and Saccade Probing
OpenMSc / BScMLHCI

See GazeTalk.AI in action

Request a demo or get in touch with our team. We welcome enquiries from speech-language therapists, hospitals, researchers, and technology partners.

etaal@dtu.dk