LAST at CMCL 2021 Shared Task: Predicting Gaze Data During Reading with a Gradient Boosting Decision Tree Approach

TL;DR

This paper presents a LightGBM-based model that predicts eye-tracking data during reading, outperforming deep learning systems in the 2021 CMCL Shared Task by using lexical, psychometric, and bigram features.

Contribution

The paper introduces a gradient boosting decision tree approach that effectively predicts gaze data, surpassing deep learning methods in the shared task.

Findings

Achieved top performance on two eye-tracking measures.

Outperformed all deep-learning based systems in the challenge.

Ranked first on the official challenge criterion.

Abstract

A LightGBM model fed with target word lexical characteristics and features obtained from word frequency lists, psychometric data and bigram association measures has been optimized for the 2021 CMCL Shared Task on Eye-Tracking Data Prediction. It obtained the best performance of all teams on two of the five eye-tracking measures to predict, allowing it to rank first on the official challenge criterion and to outperform all deep-learning based systems participating in the challenge.