Discovering Interesting Relationships among Deep Web Databases: A Source-Biased Approach
Discovering Interesting Relationships among Deep Web Databases: A Source-Biased Approach
复制标题
DOI：
10.1007/s11280-006-0227-7
复制
发表时间：
2006-12
期刊：
影响因子：
--
通讯作者：
James Caverlee;Ling Liu;D. Rocco
中科院分区：
文献类型：
--
作者：
James Caverlee;Ling Liu;D. Rocco
在过去的十年中，深层网络数据库的升级非常惊人，引发了人们对自动发现可用深层网络数据库之间有趣关系的日益浓厚的兴趣。与静态页面的“表面”网络不同，这些深层网络数据库通过基于网络的查询界面提供数据，并占所有网络内容的很大一部分。本文提出了一种新颖的偏向源的方法，可以有效地发现深网上支持网络的数据库之间的有趣关系。我们的方法通过基于源的数据库分析和探索，支持对深层网络数据库集合的以关系为中心的视图。我们的偏向源的方法具有三个独特的功能：首先，我们开发了偏向源的探测技术，该技术使我们能够通过非常精确的探针探测目标，在很少的交互中确定目标数据库是否与源数据库相关。其次，我们引入偏源相关性度量来评估所发现的深网数据库的相关性，识别深网数据库集合中有趣的偏源关系类型，并相应地对它们进行排名。发现的偏向源的关系不仅为每个深层网络数据库提供增值元数据，而且还可以为以关系为中心的个性化查询提供直接支持。第三，但并非最不重要的是，我们还使用带有焦点项的源偏置探测来开发性能优化，以进一步提高基本源偏置模型的有效性。原型系统旨在使用偏向源的方法在深层网络数据库上进行爬行、探测和支持以关系为中心的查询。我们的实验评估了所提出的源偏向分析和发现模型的有效性，表明源偏向方法优于查询偏向探测和无偏探测。
The escalation of deep web databases has been phenomenal over the last decade, spawning a growing interest in automated discovery of interesting relationships among available deep web databases. Unlike the “surface” web of static pages, these deep web databases provide data through a web-based query interface and account for a huge portion of all web content. This paper presents a novel source-biased approach to efficiently discover interesting relationships among web-enabled databases on the deep web. Our approach supports a relationship-centric view over a collection of deep web databases through source-biased database analysis and exploration. Our source-biased approach has three unique features: First, we develop source-biased probing techniques, which allow us to determine in very few interactions whether a target database is relevant to the source database by probing the target with very precise probes. Second, we introduce source-biased relevance metrics to evaluate the relevance of deep web databases discovered, to identify interesting types of source-biased relationships for a collection of deep web databases, and to rank them accordingly. The source-biased relationships discovered not only present value-added metadata for each deep web database but can also provide direct support for personalized relationship-centric queries. Third, but not least, we also develop a performance optimization using source-biased probing with focal terms to further improve the effectiveness of the basic source-biased model. A prototype system is designed for crawling, probing, and supporting relationship-centric queries over deep web databases using the source-biased approach. Our experiments evaluate the effectiveness of the proposed source-biased analysis and discovery model, showing that the source-biased approach outperforms query-biased probing and unbiased probing.