�ݺ�ߣ

1. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions A Method for Filtering Large Conceptual Schemas Antonio Villegas and Antoni Olivé {avillegas, olive}@essi.upc.edu Services and Information Systems Engineering Department Universitat Politècnica de Catalunya A. Villegas and A. Olivé ER 2010 November 3, 2010 1 / 28

2. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Outline 1 Introduction 2 Filtering Method 3 Filtered Conceptual Schema 4 Experimentation 5 Conclusions A. Villegas and A. Oliv´e ER 2010 November 3, 2010 2 / 28

3. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conceptual Schemas A. Villegas and A. Oliv´e ER 2010 November 3, 2010 3 / 28

10. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conceptual Schemas osCommerce A. Villegas and A. Oliv´e ER 2010 November 3, 2010 4 / 28 84 Entity types, 209 Attributes, 183 Relationship types, 28 IsA Relationships, 204 general constraints and derivation rules

11. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conceptual Schemas Health Level 7 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 5 / 28 2,695 Entity types, 160 Attributes, 228 Relationship types, 2,934 IsA Relationships

12. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conceptual Schemas ResearchCyc A. Villegas and A. Oliv´e ER 2010 November 3, 2010 6 / 28 26,725 Entity types, 1,060 Attributes, 5,514 Relationship types, 43,323 IsA Relationships

13. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conceptual Schemas Usability Problem The largeness of conceptual schemas makes it diﬃcult for a user to get the knowledge of interest to her This task needs computer support A. Villegas and A. Oliv´e ER 2010 November 3, 2010 7 / 28

17. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions General Overview Filtering Method Overview The main idea is to extract a reduced and self-contained view from the large schema, that is, a ﬁltered conceptual schema with the knowledge of interest to the user. A. Villegas and A. Oliv´e ER 2010 November 3, 2010 8 / 28

18. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Concrete Overview Filtering Method Overview - A user focuses on one o more entity types of interest - The method ﬁlters the large schema obtaining a subset of relevant elements to the user, taking into account: the importance of each entity type in the whole schema, and its closeness to the entity types in the user focus. A. Villegas and A. Oliv´e ER 2010 November 3, 2010 9 / 28

19. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Representing the User Request The information need of a user looking for (a subset of) the knowledge represented in a large schema includes: Focus Set What the user is interested in about the schema Rejection Set What the user is not interested in about the schema Filter Size How much knowledge the user wants to obtain from the schema at a given moment A. Villegas and A. Oliv´e ER 2010 November 3, 2010 10 / 28 Knowledge Request

23. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Representing the User Request Example of Knowledge Request The Knowledge Request is based on the concept of Entity Type. The user wants to know information about taxes in the osCommerce schema: Focus Set FS = {TaxRate, TaxClass} Rejection Set RS = {Language} Filter Size K = 10 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 11 / 28 Knowledge Request

27. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Closeness of Entity Types (Ω(e, FS)) Our method uses the closeness Ω(e, FS) between each candidate entity type e in the schema with respect to the entity types of the focus set FS. We say that e is a candidate entity type if e /∈ FS ∪ RS. Intuitively, the closeness of candidate e should be directly related to the inverse of the distance of e to the focus set FS, Ω(e, FS) = |FS| X e ∈FS d(e, e ) A. Villegas and A. Oliv´e ER 2010 November 3, 2010 12 / 28 Closeness

28. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Closeness of Entity Types (Ω(e, FS)) Example Ω(C, FS = {A, B}) Ω(e, FS) = |FS| X e ∈FS d(e, e ) Ω(C, FS) = |{A, B}| d(C, A) + d(C, B) Ω(C, FS) = 2 3 + 2 = 0.4 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 13 / 28 Closeness

29. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Closeness of Entity Types (Ω(e, FS)) Example Ω(C, FS = {A, B}) Ω(e, FS) = |FS| X e ∈FS d(e, e ) Ω(C, FS) = |{A, B}| d(C, A) + d(C, B) Ω(C, FS) = 2 3 + 2 = 0.4 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 13 / 28 Closeness

30. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Importance of Entity Types (Ψ) The importance Ψ(e) of an entity type e ∈ E of a conceptual schema is a real number that measures the relevance of e in the schema. There are several methods: Occurrence counting Ψ(e) depends on the number of characteristics the schema has about e Link analysis Ψ(e) depends on the importance of those entity types connected to e Instance-dependent Ψ(e) depends on the instances of e Our ﬁltering method can be used in connection with any of the existing importance-computing methods. A. Villegas and A. Oliv´e ER 2010 November 3, 2010 14 / 28 Importance

35. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Importance of Entity Types (Ψ) osCommerce Ψ(e) Product 1 Order 0.63 Language 0.53 Customer 0.39 Store 0.35 OrderLine 0.34 Zone 0.34 TaxZone 0.29 Special 0.28 Country 0.26 HL7 Ψ(e) Act 1 Role 0.68 ActRelationship 0.53 Participation 0.49 Entity 0.46 Observation 0.35 InfrastructureRoot 0.24 Organization 0.23 RoleLink 0.21 FinancialTransaction 0.2 ResearchCyc Ψ(e) Individual 1 LexicalWord 0.13 PartiallyTangible 0.12 Thing 0.07 SpatialThing 0.06 Agent 0.05 Organization 0.05 SomethingExisting 0.05 TemporalThing 0.04 HumanActivity 0.04 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 15 / 28 Importance

36. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Importance of Entity Types (Ψ(e)) Closeness of Entity Types (Ω(e, FS)) Interest of Entity Types (Φ(e, FS)) Φ(e, FS) = α × Ψ(e) + (1 − α) × Ω(e, FS) A. Villegas and A. Oliv´e ER 2010 November 3, 2010 16 / 28 Interest

39. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Interest of Entity Types (Φ(e, FS)) Φ(e, FS) = α × Ψ(e) + (1 − α) × Ω(e, FS) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Importance Ψ ClosenessΩ r r A. Villegas and A. Oliv´e ER 2010 November 3, 2010 17 / 28 Interest

40. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtering Measures Interest of Entity Types (Φ(e, FS)) Φ(e, FS) = α × Ψ(e) + (1 − α) × Ω(e, FS) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 P1 P2 P3P4 P5 P6P7 P8 Importance Ψ ClosenessΩ r dist(P6,r) dist(P2,r) A. Villegas and A. Oliv´e ER 2010 November 3, 2010 17 / 28 Interest

41. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Conceptual Schema (FCS) Main Task Construct a filtered conceptual schema, FCS, from the K more interesting entity types and the knowledge of the original schema CS. FCS Components - EF is a set of entity types filtered from E of CS - RF is a set of relationship types filtered from R of CS - IF is a set of IsA relationships filtered from I of CS. - CF is a set of integrity constraints filtered from C of CS. - DF is a set of derivation rules filtered from D of CS A. Villegas and A. Olivé ER 2010 November 3, 2010 18 / 28

42. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Entity Types (EF) FCS = EF , RF , IF , CF , DF A. Villegas and A. Oliv´e ER 2010 November 3, 2010 19 / 28 FS: focus set Etop: most interesting entity types Eaux : auxiliary entity types Filter Size K = |FS| + |Etop| |EF | = K + |Eaux | EF

43. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Entity Types (EF) FCS = EF , RF , IF , CF , DF A. Villegas and A. Oliv´e ER 2010 November 3, 2010 19 / 28 FS: focus set Etop: most interesting entity types Eaux : auxiliary entity types Filter Size K = |FS| + |Etop| |EF | = K + |Eaux | FS EF

44. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Entity Types (EF) FCS = EF , RF , IF , CF , DF A. Villegas and A. Oliv´e ER 2010 November 3, 2010 19 / 28 FS: focus set Etop: most interesting entity types Eaux : auxiliary entity types Filter Size K = |FS| + |Etop| |EF | = K + |Eaux | Etop FS EF

45. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Entity Types (EF) FCS = EF , RF , IF , CF , DF A. Villegas and A. Oliv´e ER 2010 November 3, 2010 19 / 28 FS: focus set Etop: most interesting entity types Eaux : auxiliary entity types Filter Size K = |FS| + |Etop| |EF | = K + |Eaux | EauxEtop FS EF

48. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Relationship Types (RF) FCS = EF , RF , IF , CF , DF The relationship types in RF are those r of the original schema whose participant entity types belong to EF or are ascendants of entity types of EF (in which case a projection of r is required) A. Villegas and A. Oliv´e ER 2010 November 3, 2010 20 / 28

51. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Relationship Types (RF) FCS = EF , RF , IF , CF , DF The relationship types in RF are those r of the original schema whose participant entity types belong to EF or are ascendants of entity types of EF (in which case a projection of r is required) A. Villegas and A. Oliv´e ER 2010 November 3, 2010 20 / 28 ← included in Eaux

52. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered IsA Relationships (IF) FCS = EF , RF , IF , CF , DF If e and e are entity types in EF and there is a direct or indirect IsA relationship between them in the original schema, then such IsA relationship must also exist in IF of FCS A. Villegas and A. Oliv´e ER 2010 November 3, 2010 21 / 28

53. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered IsA Relationships (IF) FCS = EF , RF , IF , CF , DF If e and e are entity types in EF and there is a direct or indirect IsA relationship between them in the original schema, then such IsA relationship must also exist in IF of FCS A. Villegas and A. Oliv´e ER 2010 November 3, 2010 21 / 28

54. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Filtered Conceptual Schema Filtered Integrity Constraints (CF ) and Derivation Rules (DF ) FCS = EF , RF , IF , CF , DF The integrity constraints and derivation rules included in CF and DF of FCS are those whose expressions only involve entity types from EF . Both the integrity constraint ic1 and the derivation rule dr1 are only included in CF and DF of FCS if A, B ∈ EF . A. Villegas and A. Oliv´e ER 2010 November 3, 2010 22 / 28

55. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples osCommerce A. Villegas and A. Oliv´e ER 2010 November 3, 2010 23 / 28 84 Entity types, 209 Attributes, 183 Relationship types, 28 IsA Relationships, 204 general constraints and derivation rules

56. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples osCommerce A. Villegas and A. Oliv´e ER 2010 November 3, 2010 23 / 28 FS = {TaxRate, TaxClass} and K = 10 84 Entity types, 209 Attributes, 183 Relationship types, 28 IsA Relationships, 204 general constraints and derivation rules

57. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples osCommerce A. Villegas and A. Oliv´e ER 2010 November 3, 2010 23 / 28 FS = {TaxRate, TaxClass} and K = 10 84 Entity types, 209 Attributes, 183 Relationship types, 28 IsA Relationships, 204 general constraints and derivation rules

58. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples Health Level 7 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 24 / 28 2,695 Entity types, 160 Attributes, 228 Relationship types, 2,934 IsA Relationships

59. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples Health Level 7 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 24 / 28 FS = {Patient, ActAppointment} and K = 10 2,695 Entity types, 160 Attributes, 228 Relationship types, 2,934 IsA Relationships

60. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples Health Level 7 A. Villegas and A. Oliv´e ER 2010 November 3, 2010 24 / 28 FS = {Patient, ActAppointment} and K = 10 2,695 Entity types, 160 Attributes, 228 Relationship types, 2,934 IsA Relationships

61. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples ResearchCyc A. Villegas and A. Oliv´e ER 2010 November 3, 2010 25 / 28 26,725 Entity types, 1,060 Attributes, 5,514 Relationship types, 43,323 IsA Relationships

62. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples ResearchCyc A. Villegas and A. Oliv´e ER 2010 November 3, 2010 25 / 28 FS = {Cancer} and K = 18 26,725 Entity types, 1,060 Attributes, 5,514 Relationship types, 43,323 IsA Relationships

63. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Examples ResearchCyc A. Villegas and A. Oliv´e ER 2010 November 3, 2010 25 / 28 FS = {Cancer} and K = 18 26,725 Entity types, 1,060 Attributes, 5,514 Relationship types, 43,323 IsA Relationships

64. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Conclusions The problem of filtering a fragment of the knowledge contained in a large conceptual schema appears in many information systems development activities. People needs to operate for some purpose with a fragment of the knowledge contained in those large schemas. We have proposed a filtering method based on the importance and closeness measures. A user indicates a focus set of entity types, and the method determines a subset of the elements of the large schema that is likely to be of interest to the user. We have experimented with three large schemas. In both cases, our prototype obtains the filtered schema in less than a second. A. Villegas and A. Olivé ER 2010 November 3, 2010 26 / 28

68. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions Further Work Take into account the importance of relationship types. Fine-grained ﬁltering of integrity constraints and derivation rules. Conduct experiments to precisely determine the usefulness of our method to real users. A. Villegas and A. Oliv´e ER 2010 November 3, 2010 27 / 28

71. Introduction Filtering Method Filtered Conceptual Schema Experimentation Conclusions A Method for Filtering Large Conceptual Schemas Antonio Villegas and Antoni Olivé {avillegas, olive}@essi.upc.edu Services and Information Systems Engineering Department Universitat Politècnica de Catalunya A. Villegas and A. Olivé ER 2010 November 3, 2010 28 / 28

�ݺ�ߣ

A method for filtering large conceptual schemas

More Related Content

A method for filtering large conceptual schemas