6 Conclusions
Self-integral trees were developed to answer keyword
queries in data-centric XML documents. Each mean-
ingful self-integral tree contains all or some of the
keywords to represent compact, integrated results to
the keyword queries. In addition to the content nodes
which contain at least one keyword, the meaningful
self-integral tree also contains other related nodes to
explain how the keywords are connected and how the
meaningful self-integral tree answers the keyword
query. A B+
-tree index is used to accelerate the re-
trieval of the meaningful self-integral trees in terms of
the “AND” predicate. The bloom filter is then used to
further enhance the performance of meaningful
self-integral tree retrieval, which significantly reduces
the I/O cost and achieves much higher search effi-
ciency. An effective ranking mechanism is used to im-
prove the search accuracy and the techniques of in-
dexing are examined to improve the search efficiency