Studying Visual Analytics Support for Interactive Information Retrieval within Complex Search Settings

这项研究解决了一个在我们以信息为中心的现代生活中变得越来越重要的问题。随着我们越来越熟练地在网上搜索事实信息，我们越来越多地发现需要在专门的知识库中解决复杂的信息搜索问题。虽然web搜索引擎支持使用简单的搜索界面查找焦点问题的单页答案，但对复杂信息搜索任务的支持是有限的。解决这些问题需要的不仅仅是查看搜索结果列表中排名靠前的几个文档；搜索者可能需要在潜在的大量搜索结果中进行探索和理解，在信息搜索目标变得更加精确时对查询进行细化，并综合从多个查询和可能在多个信息搜索会话中收集到的信息。******本研究的目标是研究可视化分析如何在复杂的搜索设置中支持交互式信息检索活动。由此产生的“视觉搜索分析”研究将解决四个问题：(1)分析和理解执行复杂搜索任务的困难；(2)设计支持交互式信息检索和解决复杂信息搜索问题的可视化分析方法；(3)开发并迭代完善实现这些设计的原型系统；(4)研究这些方法的优点和局限性。******当一个人对搜索主题的了解有限时，考虑在数字图书馆中搜索的困难。在这种情况下，通常从一个模糊的查询开始，并在检查文档时逐渐开发有关主题的知识。视觉搜索分析方法可以使用机器学习来提取搜索结果中的关键主题，并以视觉格式呈现这些信息，以便可以看到它们之间的关系。这将使搜索者能够轻松地识别特定的兴趣，以交互方式改进查询以匹配此兴趣，并生成后续的可视化，使搜索者能够比较和理解与主题相关的搜索结果。******基于并扩展了我之前关于使用可视化来支持以人为中心的搜索的研究结果，本研究将在三个专业的在线知识库中进行可视化搜索分析的系统研究：数字图书馆、社交媒体和在线百科全书。这些存储库在文本数据和相关元数据、搜索者使用的信息搜索行为类型以及考虑搜索成功的标准方面都具有重要的差异。在多个知识库和不同的搜索行为上研究视觉搜索分析将有助于开发一个通用的视觉搜索分析框架。

This research addresses a problem that is becoming increasingly important in our modern information-centric lives. As we have become adept at searching the web for factual information, we are increasingly discovering the need to address complex information seeking problems within specialized knowledge repositories. While web search engines support finding single-page answers to focused questions using simple search interfaces, support for complex information seeking tasks is limited. Resolving such problems requires more than just viewing the top few documents in the search results list; the searcher may need to explore among and make sense of a potentially large number of search results, refine the query as the information seeking goal becomes more precise, and synthesize information gleaned from multiple queries and potentially over multiple information seeking sessions. ******The goal of this research is to study how visual analytics can support interactive information retrieval activities within complex search settings. The resulting “visual search analytics” research will address four problems: (1) analyze and understand the difficulties of undertaking complex search tasks; (2) design visual analytics approaches that support interactive information retrieval and the resolution of complex information seeking problems; (3) develop and iteratively refine prototype systems that implement these designs; and (4) study the benefits and limitations of such approaches. ******Consider the difficulty of searching within a digital library when one has limited knowledge of their search topic. In such situations, it is common to start with a vague query, and develop knowledge about the topic incrementally as documents are examined. A visual search analytics approach could use machine learning to extract key topics within the search results, and present this information in a visual format so that their relationships could be seen. This would enable the searcher to easily identify a specific interest, interactively refine the query to match this interest, and produce subsequent visualizations that enable the searcher to compare and make sense of the search results in relation to the topics.******Building upon and extending the results of my previous research on using visualization to support human-centred search, this research will undertake a systematic study of visual search analytics within three specialized online knowledge repositories: digital libraries, social media, and online encyclopedias. These repositories each feature important differences in textual data and associated metadata, the types of information seeking behaviours searchers employ, and the criteria for considering the search a success. Studying visual search analytics over multiple knowledge repositories and various searcher behaviours will enable the development of a generalized framework for visual search analytics.