Deep Web 之探讨.pptVIP

Outline Background Research Issues Related Works Our Initial Findings for Deep Web Data Integration Conclusions Background: Complex Form Structure Research Issues How much data are hidden in the Deep Web? How to identify Deep Web interfaces? How to classify the application domain of Deep Web interfaces? How to transform queries for various Deep Web interfaces? How to integrate the hidden data in the Deep Web? How to enhance semantics from integrated Deep Web data Related Works Bergman (2001) coined “Deep Web” The Deep Web contains 7500 terabytes of information, compared to 19 terabytes in the Surface Web The Deep Web contains 550 billion documents, compared to one billion in the Surface Web The Deep Web data are more focused and are of better quality 95% of the Deep Web is publicly accessible Related Works Related Works Madhavan et al. (2007) proposed a pay-as-you-go data integration architecture for Deep Web data Related Works Madhavan et al. (2008) surfaced Deep Web content and added them into a search engine index Challenges Efficiency and Scalability Selecting proper query templates How to select valid values for text input How to identify only those input combinations that generate results suitable for search engine index Related Works Vieira et al. (2009) detected multiple query templates by Clustering web pages through checking the set of distinct paths in the DOM tree of a page Generating a single template for each cluster based on a restricted bottom-up tree mapping Could be completed for a small set of pages Has a worst-case linear complexity Related Works Shestakov (2008) investigated characteristics of the Deep Web, identified and classified Deep Web interfaces, queried web databases Could handle Java Script Could handle non-HTML forms (Applets, Flash) Proposed data models for single and consecutive forms Proposed and Implemented Deep Web Query Language Deep Web查詢介面的整合 尋找Deep Web資料庫 Deep Web目錄網站覆蓋率低 網際網路具備動態的特性 並非所有的 HTML 表單都是Deep We