Bilingual language models predict second language reading influenced by first language
Bilingual Language Models Capture Cross-Linguistic Influence in Second-Language Reading
Computation and Language
Summary
Reading in a second language is often influenced by a person’s first language, but usual language models can’t capture this effect. The authors show that bilingual language models, trained on both a reader’s first and second language, better predict how people read in their second language. This improvement is strongest for readers who have an intermediate level of skill and when their first and second languages share writing styles. The work suggests that these models reflect how the first language affects reading in the second language.
What this means in practice
- •For language technology developers: Improve multilingual reading assessment tools by incorporating bilingual models that account for first language influence on second language reading behavior.
- •For cognitive computing engineers: Develop more accurate models of human reading processes in diverse language backgrounds for applications like personalized language learning or assistive reading technology.
Authors
Abderrahmane Issam, Irene E. Winther, Stefan L. Frank
Abstract
While monolingual language model (LM) surprisal is widely established as a predictor of reading times, it cannot capture the non-native (L2) reading dynamics that are shaped by the reader's first language (L1). In this work, we investigate whether the cross-lingual transfer within bilingual models aligns with human L1-driven biases during L2 reading. We evaluate bilingual models on eye-tracking data of readers from 12 diverse L1 backgrounds reading English texts. To isolate L1-specific transfer, we compare bilingual models trained on an L1 that matches the readers' L1 background to both a monolingual English baseline and bilingual models trained on other L1s. We demonstrate that matching the readers' L1 improves the prediction of L2 reading times. This predictive advantage is concentrated in late-stage eye-tracking measures, modulated by L1-L2 script overlap, and peaks among intermediate-proficiency readers. These findings provide evidence that bilingual LMs capture aspects of cross-linguistic influence, and that such influence is present in L2 reading behavior.