During the last decade, the emphasis has been on finding relevant documents or content, an objective which most of today's search and browsing techniques address. We believe during the next decade, the emphasis will shift from documents and entities to relationships-that of discovering or validating contextually relevant, meaningful and possibly complex relationships amongst the entities that documents mention and describe. This also is likely to be the next area of focus for the Semantic Web community after it makes progress in addressing the current areas of emphasis on ontology representation, semantic metadata extraction for automatic annotation, query processing and inferencing.
Relationships are fundamental to semantics-they associate meanings to words, terms and entities. Semantic Web intends to associate annotations (i.e., metadata) with all Web-accessible resources such that programs can associate "meaning with data" to interpret them, and to process (access, invoke, utilize, and analyze) them automatically, resulting in higher scalability and better productivity. In this talk, we will look at a more comprehensive set of relationships that may be based not only on what is contained in or directly derived from data (or represented as part of resource modeling and annotations), but may be based on information extraction, external and prior knowledge and user defined computations. We also present some recent techniques for discovering indirect (i.e., transitive) and virtual (i.e., user-defined) yet meaningful (i.e., contextually relevant) relationships based on a set of patterns and paths between entities of interest, domain-specific computations and user defined relationships.
In particular, we will discuss modeling, representation and computation or validation of three types of complex semantic relationships: (a) multi-ontology relationships and the issue of "loss of information" investigated in the OBSERVER project, (b) Rho operator for semantic associations which seeks to identify contextually relevant and relevancy ranked indirect relationships or paths between entities using semantic metadata and relevant knowledge, and (c) IScapes which allows interactive, human-directed knowledge validation of hypothesis involving user-defined relationships and operations in a multi-ontology, multi-agent InfoQuilt system. A paper further describing this presentation is available at: http://lsdis.cs.uga.edu/lib/download/SAK02-TM.pdf
Amit Sheth is the director of Large Scale Distributed Information Systems (LSDIS) Lab at the University of Georgia and a Professor of Computer Science. In 1999, he founded Taalee, Inc. and managed it as its CEO until June 2001. . Since its merger with Voquette Inc. and Semagix Inc. in June 2001 he has served as CTO. Prior to joining UGA in 1994, he served at Honeywell, Unisys and Bellcore, and managed Infocosm, Inc. He is recognized for his work in federated database systems, semantic heterogeneity and semantic interoperability in distributed information systems, and workflow management. Example research projects he has led include BrAID (bridging AI-DB systems), BERDI (schema integration), InfoHarness (metadata extraction and attributed-based access of heterogeneous documents), VisualHarness (integrated multimodal-- keyword, attribute and content/feature based-- access of text and images), METEOR (multi-organizational processes and distributed workflows), CaTCH (synchronous and asynchronous collaborative teleconsulting), OBSERVER, SCORE (Semantic Content Organization and Retrieval Engine), and InfoQuilt. Commercial products resulting from this research are InfoHarness, METEOR EAppS and SCORE-based Enterprise Semantic Platform (ESP) and Knowledge Tagger. He has given 12 conference/workshop keynotes and over 120 invited/colloquium talks, and is among the best-cited authors in database/information systems literature.
xWe acknowledge contributions of current and past members of the LSDIS Lab, including Vipul Kashyap and Budak Arpinar.