Construction of Cognitive-Behavioral Model for Kana-Kanji Conversion Using Eye Gaze Information and Estimation of Concentrating on Tasks

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K17787
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61020:Human interface and interaction-related
• Research Institution: Tokyo University of Agriculture and Technology
• Principal Investigator: Tsuji Airi 東京農工大学, 工学(系)研究科(研究院), 助教 (10774284)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 視線分析 / 機械学習 / 認知負荷 / 集中

Outline of Final Research Achievements

The purpose of this study is to estimate the concentration of PC workers on their tasks in a low-impact and quantitative manner. Focusing on a highly routine cognitive judgment task, we estimated the concentration by clarifying the relationship between the worker's eye movements and cognitive resource allocation.
Discriminant models of eye movements were constructed during reading task and proofreading task based on experimental data of three conditions: 1) concentration condition, 2) a condition with a task that interferes with the main task, and 3) a condition of fatigue due to continuous work.
An F value of 0.723 was obtained for the reading task in the individual stratified 10-segment cross-validation. In addition, we attempted to personalize the discrimination model by applying individual difference standardization processing and training data selection using similarity of eye gaze behavior, and confirmed the improvement of accuracy.

Free Research Field

ヒューマンインタフェース

Academic Significance and Societal Importance of the Research Achievements

知的生産性の向上は労働人口減少が進む日本社会全体の喫緊の課題であり、様々な社会的取り組みが行われている。作業者の集中度の維持が生産性向上させる上での大きな課題となっており、その支援のためには集中度の定量的な推定が必要不可欠である。本研究では日常的な作業中に偏在し定型性の高い認知判断タスクである「黙読」と「文章入力」に着目し、作業者の視線運動と認知資源配分すなわち集中度との関係性を実験結果に基づきモデル化することで、集中度を推定する。また、個人データのみならず個人差標準化処理および類似度を用いたデータ選抜による汎化性能向上についても検討し、実装する。