Research On Behavior Analysis And Prediction Of Academic Information Retrieval Based On Eye-Tracking And Deep Learning

Studying the behavior of information seekers and their interaction with retrieval systems has always been one of the important topics in the field of information retrieval.In recent years,eye tracking technology has become an increasingly popular method in interactive information retrieval research because of its objectivity and real-time characteristics.Eye tracking technology can record the user's behavior process in real time,reflect the user's cognition,and objectively reproduce the user's visual behavior process during browsing,retrieval and other behaviors.Therefore,it is an important topic to select measurement tools and collaborative eye tracking technology to reflect users' deep-seated information retrieval behavior and improve the algorithm accuracy of users' information retrieval behavior prediction.In order to make up for the research gap in this field,on the one hand,this thesis comprehensively and clearly analyzes the changes of users' behavior,cognition during academic information retrieval by using eye tracking technology for collaborative analysis of multimodal data;On the other hand,an improved eye movement data classification algorithm based on attention mechanism is proposed to improve the accuracy of the model in predicting user information retrieval behavior.The main contents of the study are as follows:1.This thesis designs an academic retrieval behavior research experiment based on eye movement tracking.Using the knowledge of physiology,neuroscience and computer science,this thesis compares the behavioral differences of subjects in different retrieval systems and different types of tasks.After analysis,it is found that there is no significant difference in subjects' gaze in Google academic and Microsoft academic systems,the retrieval system was not the main reason for the differences of retrieval behavior.There are significant differences in the fixation data of subjects in the types of retrieval tasks.The fixation times and time of subjects in academic literature tasks are significantly less than those in academic entity tasks and academic exploration tasks.The subjects in academic exploration tasks focus on more data and more complex data.The type of the task is a significant reason for the success or failure of subjects' retrieval.2.In order to improve the accuracy of information retrieval intention prediction,an improved eye movement data classification algorithm based on attention mechanism(CNN-BLSTM-f-a)is designed in this thesis.The algorithm not only introduces the focal loss function to strengthen the learning of difficult samples,but also adds the attention module to the network structure to strengthen the important features.Through these improvements,the classification ability of the model is improved.The performance of the improved algorithm is compared with the support vector machine,naive Bayes,random forest and logistic regression methods in the traditional machine learning algorithm.The experimental results show that the prediction accuracy of the improved algorithm is higher than that of the traditional classification prediction algorithm.Figure [17] table [13] reference [102]...