基于眼动数据的网络搜索行为预测方法

预测用户的网络搜索行为对改进搜索引擎和提升用户体验十分重要.现有大多数方法是基于用户的交互数据,如查询、点击和鼠标移动等.提出一种基于眼动数据的用户网络搜索行为预测方法.通过眼动实验,采集用户在网络搜索任务时的眼睛运动数据,将这些数据转化成两种数据格式:直方图和序列.直方图数据描述用户注意力的分布情况,序列数据呈现用户的扫视路径.使用4种学习算法对用户决策或用户意图进行预测,同时研究两种数据格式的性能.结果显示,两种数据格式均适合于预测用户决策,而序列数据更适合于预测用户意图.该结果表明,利用眼动数据能够有效预测网络搜索行为.

Predicting web search behavior based on gaze data

Predicting user's web search behavior is important for search engine improvements and user experience enhancements. Most existing studies are based on user's interaction data, including queries, clicks, cursor movements, etc. The prediction of web search behavior using gaze data was presented. To capture user's eye movement data during web search tasks, an eye-tracking study was conducted and the data were transformed into two data formats: histogram and sequence. The histogram data describes the distribution of user's attention while the sequence data presents user's scan path. Four learning methods were used to predict user decision or user intent and investigate the performances of the two data formats at the same time. The results show that the two data formats are both suitable for user decision prediction, and the sequence data is more suitable for user intent prediction. The results also demonstrate the effectiveness of web search behavior prediction using gaze data.