Keep Eyes on the Sentence: An Interactive Sentence Simplification System for English Learners Based on Eye Tracking and Large Language Models
Taichi Higasa, Keitaro Tanaka, Qi Feng, Shigeo Morishima
Abstract
Language learners should read challenging texts regularly. However, using dictionaries or search engines to look up difficult expressions can be time-consuming and distracting. To address this, we have developed a system combining eye tracking with Large Language Models (LLMs) to simplify sentences automatically, allowing learners to focus on the content. The system incorporates user-tailored models that estimate users’ comprehension of sentences using gaze data and sentence information. The system also features user-triggered simplification, resulting from iterative design improvements. We conducted a user study with 17 English learners where they read English text using either our system or a baseline involving online dictionaries and search engines. Our system significantly improved both reading speed and comprehension, especially for complex sentences. The gaze-based simplification improved concentration on the content, allowing for an interruption-free reading experience. It could assist in daily reading practice, particularly for extensive reading focused on large volumes of text at a rapid pace.