Capturing Semantics of Semi-Structured Data using Partial-Order Trees

K. Goczyla (Poland)


Semi-structured data, classification, partial order, PO+tree, indexing.


The paper presents a new approach to the problem of classification of semi-structured data into abstraction classes. The criteria for classification of data are based on their embedded structure (called type) and on their position in the data graph (called role). To this aim we exploit the Object Exchange Model that represents a database of semi-structured data as a directed graph with labelled arcs. Based on this representation, we define types and roles of semi-structured objects and organise them into tree structures that capture similarities between data. The similarities are inferred from the partial order imposed on types and roles. The trees we use are partial order (PO+) trees proposed elsewhere for indexing non relational databases on multi-valued attributes. Here, PO+-trees are used to index a database of semi-structured data objects on their types and roles. As a result we obtain two layers of classification of semi-structured data.

Important Links:

Go Back