A screening process was investigated using tree-structured
data as searching files. This program is brought about by
atom codes in the tree files. No additional screening process
is necessary for ring information, because the atom codes
include ring attributes of the nodes. Two screening methods
were examined in terms of the screening procedure. In the
single-key screening, the prescreening process selects the key
which gives the smallest sets of candidates and is followed
by the screening process with the selected key. On the other
hand, the all-key screening is not so effective to reduce the
candidates despite the longer searching time. Although both
of the methods show excellent screening efficiencies and
short search times, the prescreening method has the advantage
of short search times. Comparison between the fragment-
oriented screening and the tree-structured screening
indicates that the latter system is much more improved in
both searching time and screening efficiency. And it also
shows excellent features on the exact matching retrieval. This
suggests the advantage of tree-structured data