A systematic review of eye-tracking data in NLP: exploring low-cost and cross-lingual possibilities
Nelda Kote, Hakik Paci, Alba Haveriku, Paola Shasivari, Elinda Kajo Meçe · International Journal of Grid and Utility Computing · 2024
Integrating eye-tracking data into text processing models is consistently demonstrating improvements in their outcomes.Numerous studies have been undertaken to explore low-cost alternatives and investigate cross-lingual possibilities.In our systematic literature review, we provide an overview of the related studies, based on four main dimensions: eye-tracking in Natural Language Processing (NLP) subfields, cross-lingual eye-tracking, most relevant eyetracking devices and low-cost eye-tracking opportunities.We highlight key studies showcasing that integrating eye-tracking data during training or testing improves the accuracy of NLP models in diverse subfields.There is a necessity to analyse eye-tracking data across different languages to explore cross-lingual patterns and variations.Furthermore, eye-tracking devices vary in form, sampling rate, accuracy and costs.Notably, low-cost devices are demonstrating acceptable accuracy rates, paving the way for a potentially cost-effective future in conducting eye-tracking experiments.