User Guidance and System Tuning for Retrieval Sessions

该项目旨在开发用于支持检索会议的用户指导和系统调整的模型和方法；所实施的方法将在实际业务检索服务中得到应用和评价。为此，我们将对检索会话进行定性和定量分析，以便推导出此类会话的正式描述模型。基于交互概率排序原则，这些模型将允许估计交互检索会话的质量，以及对正在进行的会话期间的有效检索策略进行预测。结合改进交互式检索方法的新实现，这些模型将用于通过提出有希望的行动，特别是更高级别的搜索活动来指导用户。为了证明新方法和模型的质量，将进行各种评估。

This project aims at the development of models and methods for user guidance and system tuning for supporting retrieval sessions; the implemented methods will be applied and evaluated in real operating retrieval services. For that, we will perform qualitative and quantitative analyses of retrieval sessions, in order to derive models for the formal description of such sessions. Based on the interactive probability ranking principle, these models will allow for estimating the quality of interactive retrieval sessions, as well as making predictions about effective retrieval strategies during ongoing sessions. Together with new implementations of methods for improving interactive retrieval, these models will be used for guiding users by proposing promising actions, especially higher level search activities. For demonstrating the quality of the new methods and models, various evaluations will be carried out.