Recommending Queries by Extracting Thematic Experiences from Complex Search Tasks.

Area of Science:

• Information Science
• Human-Computer Interaction

Background:

• Complex search tasks are often broken down into subtasks, necessitating effective query recommendation systems.
• Existing methods for subtask-oriented query recommendation rely on plain search logs (queries, clicks), offering limited information for subtask identification.
• Advancements in Computer Human Interface (CHI)/Human Computer Interaction (HCI) have led to tools generating rich search logs with more detailed user interaction data.

Purpose of the Study:

• To propose a novel method for subtask-oriented query recommendations using rich search logs.
• To leverage a visual data structure for extracting thematic experiences and identifying subtasks in complex search sessions.
• To enhance the effectiveness of query recommendations by utilizing richer user interaction data.

Main Methods:

• A visual data structure with a tree structure was proposed to log and organize rich search interactions.
• A visual-based subtask identification method was developed, utilizing the proposed data structure.
• A personalized PageRank algorithm was employed to rank nodes within identified subtasks for query recommendation.

Main Results:

• The proposed methods demonstrated effectiveness in providing subtask-oriented query recommendations.
• Experiments involving informative and tentative search tasks validated the approach.
• The use of rich search logs and a visual data structure improved subtask identification and subsequent recommendations.

Conclusions:

• Rich search logs, when structured visually, provide valuable insights for identifying subtasks in complex search.
• The developed visual-based subtask identification and personalized PageRank methods offer a promising approach for query recommendation.
• This research contributes to improving user support for complex information seeking behaviors.