Improving Retrieval of Unstructured Information using existing Information Structures

点击翻译按钮获取中文摘要

Unstructured information generally refers to text-heavy information that is not organized in a pre-defined manner such as what is found on the Web or in collections of texts. Most developments in unstructured information searching have addressed the technical algorithms for retrieving billions of Web pages; meanwhile, interfaces used for Web searching have not evolved at the same pace. To successfully use current Web searching tools, users must generally possess or acquire the vocabulary used by the authors of the relevant documents, and searchers inevitably encounter information needs that they cannot adequately express using one or two vague and broad keywords that often have multiple meanings. This is the vocabulary mismatch problem that plagues information retrieval systems—especially when searching unfamiliar knowledge domains or subject areas. This research program seeks to develop novel online information searching tools that bridge the gap between organized information (e.g., scientific, library, business or personal information collections) and Web searching. It assumes that searchers seek information to meet their needs, regardless of whether the information is structured (e.g., scientific, library or business information) or unstructured (e.g., Web searching or text). For example, structured information can suggest new keywords to better describe users’ needs. This research program capitalizes on existing information organization investments to complement unstructured information retrieval technologies. It will ensure that the tools are useful and appreciated by recording how test participants use the tools over a period of at least 3 months. Published results will include open-source online search tool prototypes, a testing engine that could be used by other interface designers and researchers, and results of the usability testing over time. This research program is innovative by virtue of the novel search systems it will design and test over time. Firstly, it will integrate existing information organization investments with the ubiquitous keyword searching and ranking in order to improve information discovery. For example, users could search a library catalogue and the Web using one integrated tool instead of two different tools they must currently use. Secondly, the tools will be tested over time to ensure that they meet searchers’ expectations and require little or no training. This type of testing over time is very rare and highly appropriate when the objective is to ensure users can truly use and appreciate a tool beyond its initial novelty. This research aims to support students who are interested in improving information exploration and searching technologies: 88% of the budget is given directly to students (i.e., salaries, travel expenses, and computers). The supported PhD and master’s students will be part of an existing research group where they will have the opportunity to collaborate with other research groups from the School of Information Studies, McGill and the University of Montreal. They will acquire skills in research, software design and development, testing, and oral/written communication, which are valuable in academic and industrial settings. Taken as a whole, this research program has the potential to improve the tools Canadian citizens use to search for all kinds of information by suggesting new keywords, grouping similar information together, and tearing down the artificial boundary between organized information collections (e.g., library catalogues or business taxonomies) and Web searching.