From Words to Texts: Using Distributional Semantics to Examine the Impact of Print-Meaning Relations at Multiple Levels on Reading Across Languages

• Applicant: Professor Dr. Sascha Schroeder
• Subject Area: General, Cognitive and Mathematical Psychology
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 561219433

In order to read efficiently, one must be able to access the meaning of written texts quickly and reliably. To achieve this, orthographic units at different linguistic levels (words, sentences, and texts) must be mapped onto corresponding semantic representations effectively. Investigating such print-meaning relations has been difficult historically because, until recently, linguistic meaning could not be quantified precisely. In recent years, however, so-called Distributional Semantic Models (DSM) have made considerable progress, and now Large Language Models (LLMs) produce high-quality vectorial representations of the meaning of words, sentences and texts. Indeed, some recent studies have used DSM-based representations to quantify the relations between orthographic and semantic units and to investigate how they affect human reading behavior. Our project builds on these studies, but will go beyond the current state of research in three important respects: First, we will use state-of-the-art DSMs to produce multiple novel metrics for print-meaning relations at the word, sentence and text levels, and experimentally investigate their effects on natural reading. Second, we will consider print-meaning relations at the three structural levels simultaneously in order to model their relative influence as well as their interactions and trade-offs. Third, we will test whether DSM-based metrics can be generalized to languages other than English and investigate the extent to which their effects are language-dependent. To do this, we will center on reading in German and Hebrew, two very different languages with different writing systems. To achieve these aims, a series of complementary studies will be carried out in four work packages that examine reading behavior in German and Hebrew using word recognition and eye-movement methods. Work packages 1-3 will each focus on the word-, sentence- and text-level and experimentally investigate the effects of the newly proposed print-meaning measures in the two languages. In work package 4, the three levels will be investigated together and the interaction of the various cues will be modeled in a new, German-Hebrew eye movement corpus. In a separate Mercator project, the relations between the new DSM-based measures and established models of text processing will be investigated. On a methodological level, the project will develop new methods for quantifying orthography and meaning at different structural levels, which apply to different writing systems. On a theoretical level, the use of these DSM-based metrics will lead to new insights into the influence of semantics on the reading process in different languages. This will better our understanding of how meaning is accessed when reading and where potential problems may arise during this process.

• DFG Programme: Research Grants
• International Connection: Israel
• Partner Organisation: The Israel Science Foundation
• Cooperation Partner: Dr. Noam Siegelman