The objective of this thesis is to automatically recognize dyslexia (reading disorder) from recorded eye-tracking data captured on the scenario of text reading. The captured raw eye-tracking data can be pre-processed into the form of fixations and saccades. Such spatio-temporal representation will be used to extract characteristic features in different forms, e.g., fixed-size vectors, time series, or images. The extracted features will be classified by suitable approaches based on nearest-neighbor and neural-network classifiers, in order to determine how much a given participant is prone to dyslexia. The proposed extraction and classification approaches will be experimentally evaluated on three text-reading tasks, for which eye-tracking data of several participants will be provided by a team at the Faculty of Arts at Masaryk University. The whole pipeline, including feature extraction and training and evaluation of various classifiers, will be implemented in Python.