Re ning OEM to Improve Features of QueryLanguages for Semistructured

Semistructured data can be modeled by the graph-oriented data model OEM. Due to its general properties, associated query languages become too complex because they work with cycles in data graphs. On the other hand, there are applications that require manipulating only tree-structured fragments of semistructured data (part-subpart relationships), and preservation of its other relationships. Therefore we reene the OEM model to distinguish between two types of edges. This reenement enables to simplify semantics of query expressions and improve default structuring of the query result by keeping a minimal structural context, which the speciied nodes had in the source data tree. We document the proposed notions on the language MailQL whose prototype is under development.