Theses

Semantic Modelling of Children's Texts

We are interested in the reading environment children grow up with. To investigate this, we collected the childLex corpus, a database of 500 children's books. We have already explored many formal properties (e.g., lexical and syntactic complexity) of these texts. However, their semantic properties are largely unexplored. In this project, you will use tools from natural language processing and topic models in order to classify the text into different thematic classes.

References: 

Schroeder, S., Würzner, K., Heister, J., Geyken, A., & Kliegl, R. (2015). childLex: A lexical database of German read by children. Behavior Research Methods, 47, 986-996.

Line Assignment of Eye Movements

We use eye-tracking in order to analyze adults' and children's reading behavior. When analyzing data from multi-line texts, it is important to ensure that fixations are assigned to the lines of the text. This is a non-trivial problem as eye movement data a noisy and often show complex non-linear distortions. In this project, you will use tools from computer vision and image registration to explore a new approach to solve this problem.

References: 

Glandorf, D. & Schroeder, S. (2021). Slice: an algorithm to assign fixations in multi-line texts. Procedia Computer Science, 192, 2971-2979.

Predicting Reading Comprehension

It has been shown that reader’s eye movements while reading texts can be used to predict their reading skill level. Most of this research in this area, however, has been conducted with native language speakers. A substantial amount of reading is conducted in a foreign language (e.g., English or French). The central aim of this project is to develop a model to predict the skill level using tools from machine learning and learning analytics.

References: 

Ahn, S., Kelton, C., Balasubramanian, A., & Zelinsky, G. (2020). Towards predicting reading comprehension from gaze behavior. ETRA, 2020.