Query Suggestion for Struggling Search by Struggling Flow Graph
Query Suggestion for Struggling Search by Struggling Flow Graph
复制标题
DOI：
10.1109/wi.2016.0040
复制
发表时间：
2016-10
期刊：
影响因子：
--
通讯作者：
Zebang Chen;Takehiro Yamamoto;Katsumi Tanaka
中科院分区：
文献类型：
--
作者：
Zebang Chen;Takehiro Yamamoto;Katsumi Tanaka
我们提出了一种生成有效查询建议的方法，旨在帮助陷入困境的搜索，即用户在搜索会话中很难找到与其信息需求相关的信息。核心是识别正在进行的挣扎会话中的挣扎成分并挖掘其有效表征。挣扎组件是用户在挣扎会话期间努力寻找有效表示的信息需求的语义组件。所提出的方法识别给定的正在进行的挣扎会话的挣扎组件，并从查询日志中挖掘包含所识别的挣扎组件的会话以构建挣扎流图。挣扎流图记录了用户对挣扎组件术语的重新表述行为，通过挣扎流图我们可以挖掘出挣扎组件的有效表示。实验结果表明，当所提出的方法可以在困难会话中使用两个或多个查询时，其性能优于基线方法。
We propose a method to generate effective query suggestions aiming to help struggling search, where users experience difficulty in locating information that is relevant to their information need in the search session. The core is identifying struggling component of an on-going struggling session and mining the effective representations of it. The struggling component is the semantic component of information need for which the user struggled to find an effective representation during the struggling session. The proposed method identifies the struggling component of given on-going struggling session and mines the sessions containing the identified struggling component from a query log to build a struggling flow graph. The struggling flow graph records users' reformulation behaviors for the terms of the struggling component, through struggling flow graph we can mine effective representations of the struggling component. The experimental results demonstrate that the proposed method outperforms the baseline methods when it can use two or more queries in a struggling session.