Looking deep in the eyes: Investigating interpretation methods for neural models on reading tasks using human eye-movement behaviour

Fariz Ikhwantri, Jan Wira Gotama Putra, Hiroaki Yamada, Takenobu Tokunaga

Abstract: Highlights•We provide a thorough analysis of NLP model interpretability with eye-movement.•Model Interpretability and eye-movement similarity are affected by several factors.•The factors include task types, interpretation methods and model architectures.•Saliency conformity between humans and machines impacts the model performance.•Our fine-grained analysis explains observations better than those in past research.

External IDs:doi:10.1016/j.ipm.2022.103195