Abstract: The eye-movements during reading have certain rules, and the reading eye-movement models help researchers better understand and recognize these rules. However, the existing reading eye-movement models have disadvantages in complex modeling approaches. This paper proposes a word-based fixation granularity processing mode and a skilled reader-based regression processing mode for reading eye-movement, making a breakthrough in the fixation granularity processing and regression of the traditional reading eye-movement models. This paper makes use of the striking parallel between reading eye-movement fixation sequence labeling and natural language sequence labeling, transforming the complex reading eye-movement modeling process into a simple language sequence labeling process, and uses maximum entropy Markov model to implement the proposed method. The experimental results show that the proposed method can better describe the reading eye-movement task, and it is easy to implement with the machine learning model.

Key words: reading eye-movement, eye-movement model, eye-tracking, sequence labeling, maximum entropy Markov

摘要： 人在阅读过程中的眼球运动具有一定规律，阅读眼动模型有助于人们更好地理解和认知这些规律。针对现有阅读眼动模型建模方法复杂的问题，突破传统阅读眼动模型注视粒度处理和回视处理模式，提出了基于单词的阅读眼动注视粒度处理模式和基于熟练读者的阅读眼动回视处理模式，利用阅读眼动注视序列标注与自然语言序列标注的高度相似性，形成了阅读眼动注视序列标注方法，从而把复杂的阅读眼动建模过程转化成了简单的语言序列标注过程，并使用最大熵马尔可夫模型实现了所提出的方法。实验结果表明，所提出的方法可以较好地描述阅读眼动任务，并且较易用机器学习模型进行实现。

关键词: 阅读眼动, 眼动模型, 眼动追踪, 序列标注, 最大熵马尔可夫