A Filter for Structured Document Retrieval

by Christian Strohamier and Holger Meuss

Structured document retrieval has established itself as a new search
area in the overlap between Database Systems and Information
Retrieval. This work proposes a filtering technique that can be added
to already existing index architectures of many structured document
retrieval systems. This new technique takes the contextual structure
information of query and document database into account and reduces
the occurrence sets handed from the index structure to the query
evaluation algorithm drastically by selecting only occurrences
appearing in the right context, and thus decreases computational
effort in query evaluation.

With the notion of "selectivity" we introduce a measure for the
added value of the filtering technique. Based on this notion, new
techniques are proposed in order to reduce space requirements for the
additional information necessary for the filtering process.  One
technique utilizes varying patterns of labels in the document database
for compressing the information. The other technique uses grammars
describing the document structure for finding labels with similar
properties. Out of a pair of those labels, only information about one
has to be included in the index structure.