US-Singapore Planning Visit: Collaborative Research on Next Generation Bibliographic Search Engine

06109998李该奖项支持一项计划访问，使位于学院公园的宾夕法尼亚州立大学的李东元博士能够与新加坡国立大学的阚敏彦博士会面。计划中的合作研究旨在建立下一代书目搜索引擎，改进集中爬行和索引、元数据提取和元数据质量等。研究人员将讨论：1)合作研究的详细范围和目标；2)协同开发的大规模元数据清洗算法的开发；3)计划车间的调度和物流。与标准搜索引擎相比，BSE （Barq搜索引擎）的主要特征是与书目参考相关的数据的独特性质和结构。这就需要一种精确的搜索，它着眼于更具体和有限的数据特征，而不是简单的关键词。示例数据将具有代表性，并且包含许多条目有缺陷的记录（缺少值、名称和标题不明确、混合引用和冗余引用）。进展将取决于与格式识别和纠错有关的新算法特性和系统级特性的发现。元数据提取是一个关键和困难的研究领域，需要国际合作努力才能取得重大进展。元数据提取是一个难点问题。对大规模元数据操作和管理的新方法的需求是迫切的。改进后的BSE系统不仅可以更有效地搜索书目文本，而且还可以用于识别和报告错误和损坏的数据，因为事先知道给定条目可能包含什么内容。改进的BSE系统还可以跨文档类型提取元数据，从而实现统一的新集合构建。这对于处理特定文档类型的社区来说是非常有益的。

0610998LeeThis award supports a planning visit to enable Dr. Dongwon Lee at the Pennsylvania State University in College Park to meet with Dr. Min-Yen Kan at the National University of Singapore. The planned collaborative research aims at building a next generation bibliographic search engine, improving focused crawling and indexing, metadata extraction, and quality of metadata, etc. The researchers will discuss: 1) the detailed scope and goals of the collaborative research; 2) the development of large-scale metadata cleaning algorithm under co-development; and 3) the scheduling and logistics of a proposed workshop. The primary characteristic of BSE (Barq Search Engine) versus standard search engines is the unique nature and structure of data associated with bibliographic references. This calls for a finely focused search that looks at data features more specific and limited than simple keywords. The sample data would be representative and contain many records with flawed entries (missing values, ambiguous names and titles, mixed and redundant citations). Progress would depend upon discovery of new algorithmic properties and system-level features associated with format recognition and error correction. Metadata extractions are a key and difficult research area and one that requires international collaborative efforts to make significant progress. The problems of metadata extraction are difficult. The need for new approaches to large-scale metadata manipulation and management are pressing. Improved BSE systems could not only search bibliographic text more efficiently, but can also serve to identify and report errors and corrupted data since there is fore knowledge of what a given entry can contain. Improved BSE systems could also extract metadata across documents genres thereby enabling uniform new collection building. This would be of great benefit to communities handling specific document types.