Schema matching Schema matching

1 schema matching

1.1 methodology
1.2 approaches
1.3 identified relationships
1.4 evaluation of quality

schema matching

methodology

discusses generic methodology task of schema integration or activities involved. according authors, 1 can view integration.

preintegration - analysis of schemas carried out before integration decide upon integration policy. governs choice of schemas integrated, order of integration, , possible assignment of preferences entire schemas or portions of schemas.
comparison of schemas - schemas analyzed , compared determine correspondences among concepts , detect possible conflicts. interschema properties may discovered while comparing schemas.
conforming schemas - once conflicts detected, effort made resolve them merging of various schemas possible.
merging , restructuring - schemas ready superimposed, giving rise intermediate integrated schema(s). intermediate results analyzed and, if necessary, restructured in order achieve several desirable qualities.

approaches

approaches schema integration can broadly classified ones exploit either schema information or schema , instance level information.

schema-level matchers consider schema information, not instance data. available information includes usual properties of schema elements, such name, description, data type, relationship types (part-of, is-a, etc.), constraints, , schema structure. working @ element (atomic elements attributes of objects) or structure level (matching combinations of elements appear in structure), these properties used identify matching elements in 2 schemas. language-based or linguistic matchers use names , text (i.e., words or sentences) find semantically similar schema elements. constraint based matchers exploit constraints contained in schemas. such constraints used define data types , value ranges, uniqueness, optionality, relationship types , cardinalities, etc. constraints in 2 input schemas matched determine similarity of schema elements.

instance-level matchers use instance-level data gather important insight contents , meaning of schema elements. these typically used in addition schema level matches in order boost confidence in match results, more when information available @ schema level insufficient. matchers @ level use linguistic , constraint based characterization of instances. example, using linguistic techniques, might possible @ dept, deptname , empname instances conclude deptname better match candidate dept empname. constraints zipcodes must 5 digits long or format of phone numbers may allow matching of such types of instance data.

hybrid matchers directly combine several matching approaches determine match candidates based on multiple criteria or information sources. of these techniques employ additional information such dictionaries, thesauri, , user-provided match or mismatch information

reusing matching information initiative has been re-use previous matching information auxiliary information future matching tasks. motivation work structures or substructures repeat, example in schemas in e-commerce domain. such reuse of previous matches needs careful choice. possible such reuse makes sense part of new schema or in domains. example, salary , income may considered identical in payroll application not in tax reporting application. there several open ended challenges in such reuse deserves further work.

sample prototypes typically, implementation of such matching techniques can classified being either rule based or learner based systems. complementary nature of these different approaches has instigated number of applications using combination of techniques depending on nature of domain or application under consideration.

identified relationships

the relationship types between objects identified @ end of matching process typically set semantics such overlap, disjointness, exclusion, equivalence, subsumption. logical encodings of these relationships mean. among others, attempt use description logics schema integration , identifying such relationships presented. several state of art matching tools today , benchmarked in ontology alignment evaluation initiative capable of identifying many such simple (1:1 / 1:n / n:1 element level matches) , complex matches (n:1 / n:m element or structure level matches) between objects.

evaluation of quality

the quality of schema matching commonly measure precision , recall. while precision measures number of correctly matched pairs out of pairs that

were matched; recall measures how many of actual pairs have been matched.