COMA A system for flexible combination of schema matching approaches. Hong-Hai Do, Erhard Rahm University of Leipzig, Germany dbs.uni-leipzig.

Transcription

1 COMA A system for flexible combination of schema matching approaches Hong-Hai Do, Erhard Rahm University of Leipzig, Germany dbs.uni-leipzig.de

2 Content Motivation The COMA approach Comprehensive matcher library Flexible combination scheme Novel reuse-oriented match approach Evaluation setup and results Conclusions and future work 2

3 Motivation Schema matching: Finding semantic correspondences between two schemas Crucial step in many applications Data integration: mediators, data warehouses E-Business: XML message mapping... Currently manual, time-consuming, tedious Need for approaches to automate the task as much as possible PO1 PO2 Customer ShipTo DeliverTo BillTo custname custstreet shiptostreet shiptocity Address custcity custzip shiptozip City Street Zip PO1.ShipTo.shipToCity PO2.DeliverTo.Address.City 3

4 Individual Match Approaches Schema-based Instance-based Reuse-oriented Element Structure Element Element Structure Linguistic Constraintbased Constraintbased Linguistic Constraintbased Dictionaries Thesauries Previous match results Names Descriptions Types Keys Parents Children Leaves IR (word frequencies, key terms) Value pattern and ranges Survey paper [Rahm, Bernstein - VLDB Journal 01] 4

5 Combining Match Approaches Combination of match algorithms Hybrid: fixed combination, difficult to extend and improve currently most common: Cupid, SemInt, SimilarityFlooding, DIKE, MOMIS, TranScm Composite: combination of the results of independently executed matchers currently only for machine learning-based techniques: LSD, GLUE COMA: Framework for flexible COmbination of MAtch algorithms Extensible matcher library Combination scheme with various combination strategies 5

6 System Architecture Matcher Library Schema Import S1 S2 UserFeedback Match Iteration Similarity cube Matcher 1 Matcher 2 Matcher 3 Matcher execution Combination of match results S1 S2 S2 S1 Mapping User Interaction (optional) Combination Scheme 6

7 Combination Scheme 1. Aggregation of matcherspecific results Average, Max, Min, Weighted SmallLarge, LargeSmall, Both 2. Match direction Matchers S1 S2 Similarity cube S1 S2 Similarity matrix S1 S2 s1 s S2 S1 s2 s Match results [S1, S2, 0.7] Combined similarity 3. Selection of match candidates MaxN (Max1), Threshold, MaxDelta, Threshold+MaxN, Threshold+MaxDelta 4. Computation of combined similarity Dice, Average 7

8 Match Processing: Example S1 shiptocity Matcher1: 0.6 Matcher2: 0.8 S2 City 1. Aggregation shiptostreet Matcher1: 0.8 Matcher2: 0.4 S1 shiptocity Average: 0.7 S2 City 2. Direction shiptostreet Average: 0.6 S1 > S2 LargeSmall (Match candidates for smaller schema S2) S2 elements City S1 elements shiptocity shiptostreet Sim Selection Max1 S2 elements City S1 elements shiptocity Sim 0.7 Threshold(0.5) S2 elements S1 elements City shiptocity City shiptostreet Sim

9 Matcher Library Type Matcher Schema Info Auxiliary Info Constituent Matchers Simple Affix Element names n-gram Element names Soundex Element names EditDistance Element names Synonym Element names External dictionaries DataType Data types Data type compatibility table UserFeedback User-specified (mis-) matches Hybrid Name Element names Affix, 3-Gram, Synonym TypeName Data Types+Names DataType, Name NamePath Names+Paths Name Children Child elements TypeName Leaves Leaf elements TypeName Reuseoriented Schema Existing schema-level match results 9

10 Reuse-oriented Matching S1 firstname lastname m S FName LName m S2 Name S1 m S2 m = MatchCompose (m1, m2) firstname lastname Name The MatchCompose operation: Transitivity of element similarity Composition of similarity relationships Reuse of multiple match correspondences vs. reuse of single element-level correspondences from synonym tables, thesauries 10

11 Schema-level Reuse The Schema matcher: Search repository S1 Si, S2 Si Match- Compose Aggregation Direction Selection S1 S2 Match problem S1 Sj, Sj S2 Sk S1, S2 Sk Existing match results Similarity cube S1 S2 Match result Reuse complete match results at the schema level Exploit all possible reuse opportunities Limit negative effects of transitivity 11

12 Real-world Evaluation 5 real-world schemas (XML Purchase order), 10 match tasks CIDX, Excel, Noris, Paragon, Apertum from biztalk.org elements Systematic evaluation (automatic mode) 1 Series = 10 Experiments: Test of 1 configuration of (Matcher, Aggregation, Direction, Selection, Combined similarity) with 10 match tasks 12,312 series = 123,120 experiments Matchers Aggregation Direction Selection Combined Sim No reuse Reuse 5 single 11 combinations 2 single 12 combinations -Max -Average -Min -Max -Average -Min -LargeSmall -SmallLarge -Both -MaxN(1-4) -Delta( ) -Threshold( ) -Threshold(0.5)+ MaxN(1-4) -Threshold(0.5)+ Delta( ) -Average -Dice -Average Σ =

13 Match Quality Measures Comparison of automatically with manually (i.e. real) derived match correspondences Real matches Suggested matches A B C A: Missed matches B: Correct matches C: False matches Quality measures: SimilarityFlooding [ICDE02]: B Precision = B + C A+ C Overall = 1 A+ B Recall = B C = A+ B B A+ B 1 = Recall* 2 Precision Overall: post-match effort to add missed and to remove false matches; negative Overall no gain Computed for single experiments and averaged over 10 experiments for each series (average Overall, etc.) 13

14 Results: Combination Strategies (1) #Series #All Series = Min Overall Most no-reuse series have negative average Overall Good matcher/strategy: Positive average Overall High presence in higher Overall ranges Series share 100% 90% 80% 70% 60% 50% 40% 30% 20% Max Average is used by all series with average Overall > 0.6 Aggregation (2376 series/strategy) Min Average 10% 0% Min Overall Aggregation: Average (compensating) 14

15 Results: Combination Strategies (2) Series share Direction: Both (considering both directions) Selection: Threshold+Delta (above threshold + within tolerance) Combined similarity: Average (pessimistic) Matcher: All (combination of all hybrid matchers) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Best selection (228 series/strategy) Thr(0.8) Delta(0.02) MaxN(1) Thr(0.5)+MaxN(1) Thr(0.5)+Delta(0.02) 0% Min Overall Series share Series share 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Sm alllarge Direction (2736 series/strategy) LargeSmall 0% Min Overall Computation of combined similarity (4104 series/strategy) 100% Dice Both Average Min Overall 15

16 Results: Single Matchers 0,9 1 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0-0,1-0,2-0,3 NamePath TypeName No reuse a) Single matchers Leaves Children Name SchemaM Reuse SchemaA avg Precision avg Recall avg Overall SchemaM: Schema with manually derived (real) match results SchemaA: Schema with match results automatically derived using the default match operation Instability of some single (hybrid) matchers (negative Overall) because of shared elements E.g. DeliverTo.Address and BillTo.Address Considering hierarchical names (NamePath) more accurate Schema-level reuse very effective: Essential improvement over no-reuse hybrid matchers Reusing approved match results better than automatically derived match results 16

17 Results: Combined Match Approaches 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 All+SchemaM SchemaM+NamePath SchemaM+Name SchemaM+TypeName b) Matcher combinations avg Precision avg Recall avg Overall SchemaM+Leaves SchemaM+Children All NamePath+Leaves NamePath+Children NamePath+TypeName NamePath+Name All: Combination of all no-reuse hybrid matchers All+SchemaM: Combination of all no-reuse hybrid matchers and SchemaM Reuse matchers outperform no-reuse matchers Best no-reuse All : 0.73 average Overall (Precision 0.95, Recall 0.78) Best reuse All+SchemaM: 0.82 average Overall (Precision 0.93, Recall 0.89) Combinations outperform single hybrid matchers Combined matchers, e.g. All, consider many aspects at the same time NamePath+Leaves: effective scheme, considering paths to identify context of shared elements, and leaves to cope with structural conflicts 17

18 Results: Match Sensitivity Overall 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 Overall(No Reuse) #All Elements 1<->2 1<->3 2<->3 1<->4 2<->4 3<->4 1<->5 2<->5 3<->5 4<->5 Match tasks Overall(Manual Reuse) Impact of schema characteristics : Degrading match quality with increase of schema size Best combinations: no-reuse All and reuse-oriented All+Schema High stability across different match tasks Little tuning effort for the default match operation # Elements 18

19 Conclusions and Future Work The COMA framework Extensible matcher library, including novel reuse approach Powerful combination scheme for both specifying match operations and constructing new matchers from existing ones Comprehensive evaluation on real-world schemas High effectiveness on large schemas Reuse: essential improvement over no-reuse Composite approach as THE solution for matcher combination Future work Matchers: more powerful reuse strategies, instance-based matchers More intelligent combination strategies Application to more real-world scenarios, esp. in bioinformatics 19