Changeset: ab84eb43b2d9 for MonetDB
URL: http://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=ab84eb43b2d9
Modified Files:
monetdb5/extras/rdf/rdfschema.c
Branch: rdf
Log Message:
implement USE_LABEL_FINDING_MAXCS (but do not enable it)
diffs (175 lines):
diff --git a/monetdb5/extras/rdf/rdfschema.c b/monetdb5/extras/rdf/rdfschema.c
--- a/monetdb5/extras/rdf/rdfschema.c
+++ b/monetdb5/extras/rdf/rdfschema.c
@@ -2970,12 +2970,68 @@ void updateParentIdxAll(CSset *freqCSset
}
}
+#if USE_LABEL_FINDING_MAXCS
+/*
+ * * Return 1 if there is semantic evidence against merging the two CS's,
this is the case iff the two CS's have a hierarchy and their common ancestor is
too generic (support above IMPORTANCE_THRESHOLD).
+ * */
+static
+char isEvidenceAgainstMerging(int freqId1, int freqId2, CSlabel* labels,
OntoUsageNode *tree) {
+ int i, j;
+ int level;
+ OntoUsageNode *tmpNode;
+
+ // Get common ancestor
+ int hCount1 = labels[freqId1].hierarchyCount;
+ int hCount2 = labels[freqId2].hierarchyCount;
+ int minCount = (hCount1 > hCount2)?hCount2:hCount1;
+
+ if (minCount == 0) {
+ // at least one CS does not have a hierarchy --> no semantic
information --> no semantic evidence against merging
+ return 0;
+ }
+
+ // get level where the hierarchies differ
+ for (i = 0; i < minCount; i++){
+ if (labels[freqId1].hierarchy[hCount1-1-i] !=
labels[freqId2].hierarchy[hCount2-1-i]) break;
+ }
+
+ if (i == 0) {
+ // not even the top level of the hierarchy is the same -->
there is semantic evidence against merging the two CS's
+ return 1;
+ } else if (i == minCount) {
+ // same name --> no semantic evidence against merging
+ return 0;
+ }
+
+ // get the common ancestor at level i
+ level = 0;
+ tmpNode = tree;
+ while(level < i){
+ for (j = 0; j < tmpNode->numChildren; j++) {
+ if (tmpNode->lstChildren[j]->uri ==
labels[freqId1].hierarchy[hCount1-1-level]){
+ tmpNode = tmpNode->lstChildren[j];
+ break;
+ }
+ }
+ level++;
+ }
+
+ if (tmpNode->percentage >= IMPORTANCE_THRESHOLD) {
+ // have common ancestor but it is too generic --> there is
semantic evidence against merging the two CS's
+ return 1;
+ } else {
+ // common ancestor is specific --> no semantic evidence against
merging
+ return 0;
+ }
+}
+#endif
+
/*
* Get the maximum frequent CSs from a CSset
* Here maximum frequent CS is a CS that there exist no other CS which
contains that CS
* */
static
-void mergeCSbyS3(CSset *freqCSset, CSlabel** labels, oid *mergeCSFreqCSMap,
int curNumMergeCS, oid **ontmetadata, int ontmetadataCount){
+void mergeCSbyS3(CSset *freqCSset, CSlabel** labels, oid *mergeCSFreqCSMap,
int curNumMergeCS, oid **ontmetadata, int ontmetadataCount, OntoUsageNode
*tree){
int numMergeCS = curNumMergeCS;
int i, j;
@@ -2983,13 +3039,12 @@ void mergeCSbyS3(CSset *freqCSset, CSlab
int tmpParentIdx;
int freqId1, freqId2;
- #if USE_LABEL_FINDING_MAXCS
- char isLabelComparable = 0;
- #endif
- char isDiffLabel = 0;
int numP1, numP2;
CS *mergecs1, *mergecs2;
- (void) labels;
+
+#if !USE_LABEL_FINDING_MAXCS
+ (void) tree;
+#endif
printf("Retrieving maximum frequent CSs: \n");
@@ -3000,44 +3055,35 @@ void mergeCSbyS3(CSset *freqCSset, CSlab
if (freqCSset->items[freqId1].type == DIMENSIONCS) continue;
#endif
- #if USE_LABEL_FINDING_MAXCS
- isLabelComparable = 0;
- if ((*labels)[i].name != BUN_NONE) isLabelComparable = 1; // no
"DUMMY"
- #endif
-
for (j = (i+1); j < numMergeCS; j++){
freqId2 = mergeCSFreqCSMap[j];
#if NOT_MERGE_DIMENSIONCS
if (freqCSset->items[freqId2].type == DIMENSIONCS)
continue;
#endif
- isDiffLabel = 0;
- #if USE_LABEL_FINDING_MAXCS
- if (isLabelComparable == 0 ||
strcmp((*labels)[freqId1].name, (*labels)[freqId2].name) != 0) {
- isDiffLabel = 1;
- }
- #endif
-
- if (isDiffLabel == 0){
- numP2 = freqCSset->items[freqId2].numProp;
- numP1 = freqCSset->items[freqId1].numProp;
- if (numP2 > numP1 && (numP2-numP1)<
MAX_SUB_SUPER_NUMPROP_DIF){
- if
(isSubset(freqCSset->items[freqId2].lstProp, freqCSset->items[freqId1].lstProp,
numP2,numP1) == 1) {
- /* CSj is a superset of CSi */
-
freqCSset->items[freqId1].parentFreqIdx = freqId2;
- updateLabel(S3, freqCSset,
labels, 0, freqId2, freqId1, freqId2, BUN_NONE, 0, 0, 0, ontmetadata,
ontmetadataCount, NULL, -1); // name, isType, isOntology, isFK are not used for
case CS
- break;
- }
- }
- else if (numP2 < numP1 && (numP1-numP2)<
MAX_SUB_SUPER_NUMPROP_DIF){
- if
(isSubset(freqCSset->items[freqId1].lstProp, freqCSset->items[freqId2].lstProp,
- numP1,numP2) == 1) {
- /* CSj is a subset of CSi */
-
freqCSset->items[freqId2].parentFreqIdx = freqId1;
- updateLabel(S3, freqCSset,
labels, 0, freqId1, freqId1, freqId2, BUN_NONE, 0, 0, 0, ontmetadata,
ontmetadataCount, NULL, -1); // name, isType, isOntology, isFK are not used for
case CS
- }
-
- }
+ numP2 = freqCSset->items[freqId2].numProp;
+ numP1 = freqCSset->items[freqId1].numProp;
+ if (numP2 > numP1 && (numP2-numP1)<
MAX_SUB_SUPER_NUMPROP_DIF){
+ if (isSubset(freqCSset->items[freqId2].lstProp,
freqCSset->items[freqId1].lstProp, numP2,numP1) == 1) {
+ /* CSj is a superset of CSi */
+#if USE_LABEL_FINDING_MAXCS
+ if (isEvidenceAgainstMerging(freqId1,
freqId2, *labels, tree)) continue;
+#endif
+ freqCSset->items[freqId1].parentFreqIdx
= freqId2;
+ updateLabel(S3, freqCSset, labels, 0,
freqId2, freqId1, freqId2, BUN_NONE, 0, 0, 0, ontmetadata, ontmetadataCount,
NULL, -1); // name, isType, isOntology, isFK are not used for case CS
+ break;
+ }
+ }
+ else if (numP2 < numP1 && (numP1-numP2)<
MAX_SUB_SUPER_NUMPROP_DIF){
+ if (isSubset(freqCSset->items[freqId1].lstProp,
freqCSset->items[freqId2].lstProp,
+ numP1,numP2) == 1) {
+ /* CSj is a subset of CSi */
+#if USE_LABEL_FINDING_MAXCS
+ if (isEvidenceAgainstMerging(freqId1,
freqId2, *labels, tree)) continue;
+#endif
+ freqCSset->items[freqId2].parentFreqIdx
= freqId1;
+ updateLabel(S3, freqCSset, labels, 0,
freqId1, freqId1, freqId2, BUN_NONE, 0, 0, 0, ontmetadata, ontmetadataCount,
NULL, -1); // name, isType, isOntology, isFK are not used for case CS
+ }
}
//Do not need to consider the case that the numProps
are the same
@@ -8856,12 +8902,12 @@ RDFextractCSwithTypes(int *ret, bat *sba
if (0){
/*S3: Merge two CS's having the subset-superset relationship */
- mergeCSbyS3(freqCSset, labels, mergeCSFreqCSMap,curNumMergeCS,
ontmetadata, ontmetadataCount);
+ mergeCSbyS3(freqCSset, labels, mergeCSFreqCSMap,curNumMergeCS,
ontmetadata, ontmetadataCount, ontoUsageTree);
curNumMergeCS = countNumberMergeCS(freqCSset);
curT = clock();
- printf("Merging with S4 took %f. (Number of mergeCS: %d | NumconsistOf:
%d) \n", ((float)(curT - tmpLastT))/CLOCKS_PER_SEC, curNumMergeCS,
countNumberConsistOfCS(freqCSset));
- printf("Number of added CS after S4: %d \n", freqCSset->numCSadded);
+ printf("Merging with S3 took %f. (Number of mergeCS: %d | NumconsistOf:
%d) \n", ((float)(curT - tmpLastT))/CLOCKS_PER_SEC, curNumMergeCS,
countNumberConsistOfCS(freqCSset));
+ printf("Number of added CS after S3: %d \n", freqCSset->numCSadded);
#if STORE_PERFORMANCE_METRIC_INFO
computeMetricsQ(freqCSset);
_______________________________________________
checkin-list mailing list
checkin-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/checkin-list
