ETDD70: Eye-Tracking Dataset for Classification of Dyslexia using AI-based Methods

SEDMIDUBSKÝ, Jan, Nicol DOSTÁLOVÁ, Roman ŠVAŘÍČEK and Wolf CULEMANN. ETDD70: Eye-Tracking Dataset for Classification of Dyslexia using AI-based Methods. Online. In 17th International Conference on Similarity Search and Applications (SISAP). Springer, 2024, 14 pp.

Basic information

• Original name: ETDD70: Eye-Tracking Dataset for Classification of Dyslexia using AI-based Methods
• Authors: SEDMIDUBSKÝ, Jan, Nicol DOSTÁLOVÁ, Roman ŠVAŘÍČEK and Wolf CULEMANN.
• Edition: 17th International Conference on Similarity Search and Applications (SISAP), 14 pp. 2024.
• Publisher: Springer

Other information

• Original language: English
• Type of outcome: Proceedings paper
• Field of Study: 10200 1.2 Computer and information sciences
• Confidentiality degree: is not subject to a state or trade secret
• Publication form: electronic version available online
• Organization unit: Faculty of Informatics
• Keywords in English: dyslexia;eye tracking;time-series data;classification;k-nearest neighbor query;multilayer perceptron;residual networks
• Tags: DISA
• Tags: International impact, Reviewed

Abstract

Dyslexia, a specific learning disorder, poses challenges in reading and language processing. Traditional diagnostic methods often rely on subjective assessments, leading to inaccuracies and delays in intervention. This work proposes classifying dyslexia using AI-based methods applied to eye-tracking data captured during text reading tasks. To facilitate future research in this domain, we collect a novel dataset (ETDD70) comprising eye-tracking recordings of 70 individuals for three reading tasks. In particular, the dataset contains high-frequency and accurate time series of 2D positions of eye movements and many derived characteristics extracted from eye movement patterns. By leveraging similarity-search approaches and deep learning models, we demonstrate the utility of such data in training several classification models, the best of which can distinguish between dyslexic and non-dyslexic individuals with an accuracy of around 90%. Both the dataset and evaluated models provide a valuable resource for researchers to further advance AI-based methods for dyslexia classification.

Links

• TL05000177, research and development project: Name: Diagnostika dyslexie s využitím eye-trackingu a umělé inteligence (Acronym: DYSLEX)

Investor: Technology Agency of the Czech Republic