Scholarlink

With the rapid expansion of information, it becomes a challenge to locate relevant content efficiently. Analyzing reading preferences is crucial for optimizing the reading experience and enhancing the efficiency of information acquisition. Traditional analyses of reading preferences tend to rely on individual actions and lack in-depth exploration of the reading process. These methods focus on single eye-tracking features or universal eye-tracking patterns without analyzing individual preferences and unique reading patterns. In this study, we employ a large language model to analyze text-influenced eye-tracking features and text to extract key information. By comparing the key information with the keywords of the text, factors affecting both are systematically excluded, leading to the development of analytical models to isolate reading preferences. A matching correlation of 60.4% was attained by extracting key information from text materials, demonstrating the information recognition capability of the large language model. Subsequently, key information is extracted by integrating eye-tracking and text. Segments related to Chinese texts were isolated to identify elements influenced by the participants’ reading preferences. This approach provides valuable insights to further explore cognitive mechanisms and preference formation during reading.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reading Preference Analysis Through Eye-Tracking and Large Language Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shuangshuang Ying,
  • Dongsen Zhang,
  • Huijia Wu,
  • Churan Yu,
  • Yongji Liu,
  • Zhaofeng He

摘要

With the rapid expansion of information, it becomes a challenge to locate relevant content efficiently. Analyzing reading preferences is crucial for optimizing the reading experience and enhancing the efficiency of information acquisition. Traditional analyses of reading preferences tend to rely on individual actions and lack in-depth exploration of the reading process. These methods focus on single eye-tracking features or universal eye-tracking patterns without analyzing individual preferences and unique reading patterns. In this study, we employ a large language model to analyze text-influenced eye-tracking features and text to extract key information. By comparing the key information with the keywords of the text, factors affecting both are systematically excluded, leading to the development of analytical models to isolate reading preferences. A matching correlation of 60.4% was attained by extracting key information from text materials, demonstrating the information recognition capability of the large language model. Subsequently, key information is extracted by integrating eye-tracking and text. Segments related to Chinese texts were isolated to identify elements influenced by the participants’ reading preferences. This approach provides valuable insights to further explore cognitive mechanisms and preference formation during reading.