Information Technology Reference
In-Depth Information
uses the lexical properties of archetype node text and SNOMED CT
concept descriptions to find correspondences between them. We also used nor-
malization techniques such as ignoring cases, skipping prepositions, avoiding
initials in term names, and concatenating the word “observable” at the end of
the node text. Once the candidate mapping set is recovered, a structure-based
technique is applied between the archetype node tree and the hierarchy of can-
didate SNOMED CT concepts. First, the method computes lexical alignments
between descendants of archetype root node and the narrower concepts of each
candidate SNOMED CT concept. Second, the method computes the similarity
value between candidate pairs (archetype node, SNOMED CT concept) by cal-
culating the percentage of common descendant between the two components of
the pair. Then, the method applies a conservative solution of the problem: it
selects the mapping with the highest similarity value, rejecting the rest.
Algorithm 2. Archetype Root Node Mapping using SNOMED CT
MapRootNode (Node n)
begin
1
snomedCandidateConcepts ←− MapObservableConceptByName( n.text )
2
if
|snomedCandidateConcepts| =0 then
3
snomedCandidateConcepts ←−
4
MapObservableConceptByName( n.text +“observable” )
rootDescendants ←− GetElementDescendantNodes( n )
5
maxChildMapping ←− 0
6
for
candidate ∈ snomedCandidateConcepts
do
7
candidateChildren ←− GetChildrenOfConcept( candidate )
8
mappedChildrenCount ←−
0
9
for descendant ∈ rootDescendants do
10
generatedNames ←− CombineNames( descendant.text, n.text )
11
for child ∈ candidateChildren do
12
for name ∈ generatedNames do
13
if Map( child,name ) then
14
mappedChildrenCount ←− mappedChildrenCount +1
15
if mappedChildrenCount > maxMappedChildren then
16
winners ←− candidate
17
maxMappedChildren ←− mappedChildrenCount
18
if mappedChildrenCount = maxMappedChildren then
19
winners. add( candidate )
20
return winners
21
3.3 Mapping of the Element Nodes
Direct Mapping to SNOMED CT. All archetypes element node that have
not been mapped to some SNOMED CT concept in Algorithm 2 are mapped in
this stage. For each SNOMED CT candidate retrieved using the
previous lexical matching, the method (Algorithm 3) traverses the relation in-
terprets searching possible candidates for the values of the original element
node (Fig. 3). After Algorithm 3, the non-mapped element nodes prospect for
SNOMED CT correspondences using the previous “named-based” technique.
 
Search WWH ::




Custom Search