Integration of Heterogeneous Information Sources

This thesis presents a framework for integrating information from multiple heterogeneous sources. The framework consists of an HTML extension called OHTML and the W3Graph query language. OHTML adds to HTML semantic tags. The W3Graph query language is a query language for semistructured data that we designed for querying OHTML pages. We designed OHTML to enrich HTML pages with semantic information describing the contents of the pages. The OHTML format facilitates expressing the metadata (information about information) by providing a mechanism for embeding semantic tags in HTML pages. These semantic tags describe OEM data structures presented in HTML documents. The data structures do not have an absolute schema xed in advance. This is the reason why the information described in such a way is called semistructured. Formally, the data structure corresponds to the Object-Exchange Model (OEM) described in [PGMW95]. We present the W3Graph query language for querying OHTML pages. The queries in our language are de ned by means of graphs. Such approach provides the important advantages of conciseness and clarity, espessialy when de ning queries that involve several path expressions. Another important feature of the W3Graph query language is that it allows restructuring of query results according to user needs and producing answers as sets of OHTML documents. We de ne the notion of a maximal assignment and we divide constraints in two di erent ways. First, constraints are either weak or strong, depending on how their semantics is de ned in the presence of nulls. Second, constraints are either search or lter constraints, depending on how they are used to form the result of a query. Enriching the language with these constructions allows us to handle e eciently and in a conceptually convenient way missing or incomplete information. The role of maximal assignments is very close to the role of full disjunctions de ned in [GL94]. We describe an algorithm for computing maximal assignments. The running time of the algorithm is polynomial in the size of the input the output. The thesis describes also the W3Graph system. The system implements the W3Graph query language and the query-processing algorithm. The system provides users with querying OHTML pages and publishing restructured results in OHTML format. The system has been written entirely in Java as a combination of a server and a client. The client runs as an applet in a Web browser. ii