Modeling Bilingual Sentence Processing: Evaluating RNN and Transformer Architectures for Cross-Language Structural Priming
Published in Proceedings of the Fourth Workshop on Multilingual Representation Learning (MRL 2024), 2024
Recommended citation: Zhang, D., Xiao, B., Gao, C., Youm, S., & Dorr, B. (2024). Modeling Bilingual Sentence Processing: Evaluating RNN and Transformer Architectures for Cross-Language Structural Priming. In Proceedings of the Fourth Workshop on Multilingual Representation Learning (MRL 2024). https://aclanthology.org/2024.mrl-1.8/
This study evaluates the performance of Recurrent Neural Network (RNN) and Transformer models in replicating cross-language structural priming, a cognitive phenomenon where prior exposure to a sentence structure increases the likelihood of selecting a similar one later. Focusing on Chinese-English priming, the authors found that Transformer models outperform RNNs in generating primed structures, with accuracy ranging from 25.84% to 33%. These findings challenge the idea that human sentence processing is purely recurrent and suggest a cue-based mechanism better reflected by Transformer behavior.