CONCURRENCY AND COMPUTATION: PRACTICE AND EXPERIENCE Concurrency Computat.: Pract. Exper. 2011; 23:956–971 Published online 28 September 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpe.1624 Automatically constructing semantic link network on documents Hai Zhuge 1,2, ∗, † and Junsheng Zhang 1 1 Knowledge Grid Research Group, Key Lab of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, 100190 Beijing, People’s Republic of China 2 Southwest University, People’s Republic of China SUMMARY Knowing semantic links among resources is the basis of realizing machine intelligence over large-scale resources. Discovering semantic links among resources with limited human interference is a challenge issue. This paper proposes an approach to automatically discovering and predicting semantic links in a document set based on a model of document semantic link network (SLN). The approach has the following advantages: it supports probabilistic relational reasoning; SLNs and the relevant rules automatically evolve; and, it can adapt to the update of the adopted techniques. The approach can support cyber space applications, such as documentation recommendation and relational queries, on large documents. Copyright  2010 John Wiley & Sons, Ltd. Received 6 May 2010; Accepted 16 May 2010 KEY WORDS: semantic link network; probability; rules; relational reasoning; inference 1. INTRODUCTION 1.1. Motivation Knowing relations between resources is important for realizing intelligent applications on large- scale resources. Manually establishing relations is time-consuming; hence, it is necessary to study the approach to automatically discover relations among resources. Rethinking the success of the World Wide Web indicates the right way to the future semantic Web: inheriting the features of the Web—the simple hyperlink mechanism and the easy utility mode. The World Wide Web is a network of references implemented by hyperlink. The content of a web page is explained by its text and the reference pages. The hyperlink represents the relevant topics of the page, whereas it does not represent the relation between pages explicitly. Therefore, the hyperlinked Web cannot recommend useful web pages. The semantic link network (SLN) model extends the Web by attaching semantic indicators to hyperlinks [1]. A typical SLN consists of semantic nodes, semantic links and relational reasoning ∗ Correspondence to: Hai Zhuge, China Knowledge Grid Research Group, Key Lab of Intelligent Information Processing, Institute of Computing Technology, P. O. Box 2704-28, 100190 Beijing, People’s Republic of China. † E-mail: zhuge@ict.ac.cn Contract/grant sponsor: National Basic Research and Development Program; contract/grant number: 2003CB317001 Contract/grant sponsor: International Cooperation Project of Ministry of Science and Technology of China; contract/grant number: 2006DFA11970 Contract/grant sponsor: National High Technology Research and Development Program of China; contract/grant number: 2007AA12Z220 Contract/grant sponsor: National Science Foundation of China; contract/grant numbers: 60773057, 60703018 Copyright  2010 John Wiley & Sons, Ltd.