Transforming extended entity-Relationship model into owl ontology in temporal databases

Journal of Computer Science and Cybernetics, V.34, N.1 (2018), 77–96 DOI 10.15625/1813-9663/34/1/10110 TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL INTO OWL ONTOLOGY IN TEMPORAL DATABASES VO HOANG LIEN MINH1, QUANG HOANG Hue University of Sciences, Hue University E-mail: 1minhvhl@gmail.com
Abstract. The objective of W3C is to transform the current Web into a Semantic Web to reuse the previous system. Many previous systems were built based on the ER model, so the upgradation and transformation of the ER model into ontology in order to reduce cost is really necessary. There are various studies aiming at transforming from ER and EER model into ontology; however, the studies have not still mentioned the transforming from the temporal databases into OWL ontology. Therefore, this paper proposes the rules for transforming the temporal components in TimeER model into OWL. Keywords. Conceptual data model; ontology; OWL; recursive relationship; Semantic Web; Ti- meER model. 1. INTRODUCTION In recent years, ontology has become a well-known term in the field of computer science for helping the machine “understand” and so that they could be able to process the information efficiently. Basically, ontology provides vocabulary to describe data which helps the computer understand their semantic. There are many languages designed to present ontology for semantic web. Among those ones, Resource Description Framework Schema (RDFS) and Web Ontology Language (OWL) are the two popular languages. OWL is considered as an extension of the RDFS to get rid of the disadvantages of RDFS. Nowadays, OWL is being seen as the standard language for building ontology. OWL is a language describing the classes, properties and relationships between these objects in a way that computer can understand web contents. The objective of World Wide Web Consortium (W3C) is to transform the current web into semantic web with to reuse the previous system [1]. Many previous systems were built based on the Entity - Relationship (ER) model, so upgrading and transforming the ER model into ontology to reuse the previous system and reduce the cost is really necessary. There are several studies of transformation ER model into ontology. Upadhyaya and Kumar [2] introduced the ERONTO tool with the use of Java and Jena 2.1 for extracting ontology from ER schema. M. Fahad [1] proposed a method of designing the OWL on- tology from ER model which based on a set of rules to transform the components of an ER model (entity, attribute and relationship) into the corresponding components in OWL. c© 2018 Vietnam Academy of Science & Technology 78 VO HOANG LIEN MINH, QUANG HOANG Myroshnichenko [3] introduced a solution for automatically transforming ER model into correspondence semantic in OWL Lite Ontology. In general, the studies have proposed an approach of transforming ER model into OWL ontology. However, these approaches have not formalized the transformation rules with the components of Extended Entity Relationship (EER) model (such as transformation of superclass/subclass relationship...), as well as mentioned the designing of the temporal onto- logy from the temporal ER model such as the proposed TimeER model by [4]. In addition, previous approaches have not proposed the transformation of the recursive relationship. So, this transformation into OWL ontology will affect preservation of the semantic of the ER model and its extension. Therefore, this paper will propose fully the rules for transforming the EER model into OWL ontology. Accordingly, this paper is organized as follows: the next section presents a method for transforming EER model into OWL ontology. Section 3 classifies the recursive relationship and proposes the rules for transforming them into OWL ontology. Section 4 proposes a method for transforming TimeER into OWL based on additional rules for transformation of the temporal component of TimeER model into OWL ontology. Section 5 concludes and discusses implications for future works. 2. TRANSFORMATION OF EER MODEL INTO OWL ONTOLOGY In order to illustrate the transformation of the EER model to OWL ontology and to informally validate our transforming rules, this paper transforms a real-world example of an EER model such as Figure 1 (modeling an employee management system) into an OWL ontology. Management system is used to store and manage information in a unit. The main components of the EER model are the Person, Employee, Department, Project entities with corresponding simple and composite attributes. In it, the entity Empoloyee requires that the system manages both the lifespan and transaction time (denoted LT) to record the time that an entity exists in reality and the time that the entity/event is in the database. The entity Department is required to manage transaction time. The attribute Salary is required to handle the valid time and transaction time (BiTemporal, denoted BT), to record the time that an event is true in reality and the time that the entity/event is in the database. Similarly, the multivalued attribute Location is also required to be managed on a valid time. The relationship Work for is required to handle a valid time to record the time that an event is true in reality. Figure 18 is the result of applying transformation rules. Ontology consists of classes corresponding to the entities of the ER model and object properties corresponding to ER relationship. All of the cardinality constraints in the ER model have been correctly trans- formed to equivalent OWL restriction. 2.1. Transformation of entities type In the ER model, entities are used to describe complex constructs such as people, pla- ces, things or events of interest. In OWL, the class provides an abstract mechanism for representing objects with similar properties. We have a transformation rule as follows: TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 79 Figure 1. An example of TimeER model Rule EER1. Transform each entity type E into class C (E ) by OWL ontology [5]. Figure 2. Transformation of entity type and attribute In Figure 2, the entity Employee was transformed into class Employee in OWL ontology. 2.2. Transformation of attributes In OWL, the datatype attribute is a single valued property, so they are equivalent to the single valued attribute in the ER model. And to define a single valued property and have only one value, set it with a cardinality constraint. In the ER model, the primary key is used to uniquely identify an entity. It aims to ensure the condition of a single attribute with 80 VO HOANG LIEN MINH, QUANG HOANG a single value, and it is impossible to be a null and must be distinct value. To represent the primary key without getting a null value in OWL, set minCardinality of the attribute to 1 and maxCardinality to 1, meaning that the cardinality is 1. We have a transformation rule as follows: Rule EER2. The single valued attribute attE of the entity E was transformed into the datatype property attE with range is corresponding with data type in OWL, and domain is E, set function characteristics for datatype property attE [4]. If the attribute attE is not NULL then set minimum cardinality constraint to 1 for datatype property attE. For each key attribute KE of entity E, transform to datatype property KE in the set of key properties of class C (E ), set function characteristics and cardinality constraint is 1 [5]. In Figure 2, the attributes Name, Birthday, Salary will be transformed into datatype properties, with range is corresponding with data type in OWL and domain is Employee class, set function characteristics for this properties. The key attribute EmpID will be transformed to datatype property EmpID in the set of key properties of class Empolyee, domain is Employee and ranger is string . Besides, it is set function characteristics and cardinality constraint is 1 for the datatype properties EmpID by using syntax owl:cardinality in Figure 2. In the ER model, multi-valued attributes can have multiple attribute values. In OWL, by default, the OWL property allows for multiple values. Rule EER3. Transform multi valued attE of entity E similar Rule EER1, but not set function characteristics [5]. Figure 3. Transformation of multi-valued attributes In Figure 3, the multi valued attributes Location will be transformed into datatype pro- perties, with range is corresponding with data type in OWL and domain is Department class, but not set function characteristics for this properties. 2.3. Transformation of composite attribute Composite attributes are attribute that are made up of different single attributes. In OWL, it is possible to represent complex properties by setting them as child datatype pro- perties. We have a transformation rule as follows: Rules EER4. The composite attribute attE of the entity E has the subattributes sub attE will be transformed to datatype property attE of class C (E ). The subattributes sub attE will be transformed to datatype properties sub attE, and they are subproperties of datatype properties attE, with function characteristics, domain is the datatype properties atttE and range is corresponding with data type in OWL [1]. TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 81 Figure 4. Transformation of composite attribute The composite attribute Budget has two subattributes Expences and Income in Figure 4 will be transformed into datatype property Budget, and two subdatatype properties Expences and Income. 2.4. Transformation of an inheritance relationship To represent the disjointness specialization, we can use disjoint constraints between clas- ses that correspond to subentity. Rule EER5. The subentity sub E inherits from the entity E will be transformed into subclass (inheritance) in the corresponding class of the super entity. Figure 5. Transformation of an inheritance relationship In Figure 5, the entity Manager was transformed into subclass Manager in class Em- ployee, and OWL source code as follows. 2.5. Transformation of disjoint and overlap specializations Rules EER6. Transform each subentity sub Ei of a subclass in a disjoint specialization E into corresponding subclass C(sub Ei) by using syntax owl:disjointWith, set subclasses 82 VO HOANG LIEN MINH, QUANG HOANG that do not intersect with other subclass inheriting from that disjoint class E [2]. Figure 6. Transformation of disjoint specialization In consideration of the entity A, there are two disjoint classes including B and C. We have applied rule Rule EER6. in order to transform for two classes B and C by using the syntax owl:disjointWith and OWL source code as follows, shown in Figure 6. In OWL, none of the constraints has the same meaning as the overlapping in the ER model. Therefore, in order to represent this constraint, we can use the union constraint. We have a transformation rule as follows: Rule EER7. Transform the subclass sub Ei in the total completeness constraint with super class E to class C (E ) and subclass C(sub Ei). Using syntax owl:unionOf for setting those subclass C(sub Ei) [2]. 2.6. Transformation of the weak entity and identifying relationship In OWL, representation of a weak entity is similar to a strong entity, the only difference is that it adds two inverse object properties, adds the owner key properties into the set of key properties of the weak entity class. We have a transformation rule as follows: Rule EER8. Let us consider that W is a weak type of identifying relationship R and the owner entity type is E. Suppose that W has partial key KW, and KE is a key of E. A weak entity always participates in the identifying relationship with cardinality constraint (1, 1), therefore, depending on the second cardinality constraint of the identifying relationship R, we will construct the transform rules as follows: - Add the class C (W ), the attributes attW of the weak entity type W will transform into datatype properties attW of class C (W ) similar to the transformation of the attributes of the strong entity. - Add two inverse object properties EhasW and WOfE which show relationship between classes C (E ) and C (W ), their identities, ranges and domains as shown in Table 1. Set function characteristics and minimum constraint is 1 to two properties WOfE ; - Depending on the second cardinality constraint of the identifying relationship, add the corresponding min/max to the object properties EhasW (the object properties just added with domain is class corresponding with entity type which has pairs of cardinalities). - Key of class C (W ) was created by combining the partial key KW with key KE of class C (E ). TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 83 Figure 7. Transformation of the weak entity and identifying relationship Table 1. Pairs properties added when transformation of the identifying relationship Identity Domain Range EhasW C(E) C(W) WOfE C(W) C(E) As an example in Figure 7, the weak entity Dependent was transformed into class De- pendent, the indentifying relationship Dep of was transformed into pairs of inverse object properties EmployeeHasDependent and DependentOfEmployee for presenting the indenti- fying relationship. Two attributes DependentName and Relationship of the weak entity will transfomed into the datatype properties of class Dependent. 2.7. Transformation of the relationship In ER model, the relationship is how the data is shared between entities. The degree of a relationship type is the number of entity types that participate. Next section will consider three types of relationships between entities: the binary relationship, n-ary relationship and recursive relationship. 2.7.1. Transformation of the relationship without attributes The binary relationship has two cardinality constraint of the form (x, y), where x is a minimum constraint, y is maximum constraint. Typically, x is 0 or 1, and y is 1 or N [6]. Rule EER9. Let us consider the binary relationship R without attributes between two entities E1 and E2 with left cardinality constraint is (x1, y1) and right cardinality constraint is (x2, y2). We have transformation rules as follows: - Add two inverse object properties for presenting the relationship between classes C(E1) and C(E2); - With each value of cardinality constraint, if min differential 0 or max deferential N on the relationship R, add min/max restriction to the corresponding object properties. 84 VO HOANG LIEN MINH, QUANG HOANG Figure 8. Transformation of the relationship without attributes In Figure 8, the relationship Resp for between two entities Department and Project will be transformed into two inverse object properties DepartmentRespForProject and ProjectRe- spForDepartment. Because the left cardinality constraint is (1, 1), we set min and max restriction is 1 to object property ProjectRespForDepartment. 2.7.2. Transformation of the relationship with attributes A relationship can also have its own attributes (especially n − n relationships). The attributes of a relationship are single valued attributes. In this case, where the relationship R has an attribute, if R is a 1-1 relationship, then it can move it into the attribute of one of two entities, and if R is a 1-n relationship, moving this attribute into the attribute of the entity which has maximum constraint. Figure 9. Transformation of the relationship with attributes TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 85 However, if the attribute of a relationship is moved into either entity or represented in the OWL, then semantic are not guaranteed. We can treat a relationship as an entity or attribute, the designer can determine the relationship as an entity to ensure semantics during transformation. We have a transformation rule as follows: Rules EER10. Let us consider the binary relationship R with attributes attR between entities E1, E2, we have a transformation rule as follows: - Add class C (R), the attributes attR of the relationship R was transformed into the datatype properties of class C (R). - Add two inverse object properties E1HasR and ROfE1 for representing the relationship between classes C (R) and C(E1), two inverse object properties E2HasR and ROfE2 for representing the relationship between classes C (R) and C(E2). Set function characteristics and minimum constraint is 1 to two properties ROfE1, ROfE2; - With each value of min/max constraint, if min differential 0 and max differential N on the relationship R, add corresponding min/max constraint to the object properties E1HasR v E2HasR. If the binary relationship is N −N , add two object properties ROfE1, ROfE2 to the key set of properties of class R [3]. As shown in Figure 9, the relationship Belong to with attribute Joint date will be trans- formed into class BelongTo. Two inverse object properties EmployeeHasBelongTo, Belong- ToOfEmployee represent the relationship between an entity Employee and a relationship Belong To. Likewise, two inverse object properties DepartmentHasBelongTo, BelongToOf- Department represent the relationship between an entity Department and a relationship Be- long To. The attribute Joint date was transformed into the datatype properties JointDate of class BelongTo. 2.7.3. Transformation of n-ary relationship Thanks to the transforming rules of the binary relationship, we can apply those rules for the transformation of the n-ary relationship R, by considering R as an entity type, and this entity has the binary relationship with the participating entity types Ei. So these binary relationship can only be one-to-one or one-to-many binary relationship (since the cardinality constraint of the entity type R with all these binary relationships is (1, 1)). We have a transformation rule as follows: Rule EER11. Let us consider the n-ary relationship among the entity types Ei - Add class C (R), the attributes of the relationship R are transformed into the datatype properties of class C (R); - Add pairs of the inverse object properties which show the relationship between class C (R) and classes Ei, set the minimum and maximum constraint. Set 1 for min and max constraint of the inverse object properties. - Add the object properties with domain is C (R) which has just added to set of key properties of class C (R) [3]. Note: If the relationship R has function constraint, remove the object properties which has range corresponding with the entity appeared on the right of the function constraint out of the set of key properties of class C (R). 86 VO HOANG LIEN MINH, QUANG HOANG 3. TRANSFORMATION OF THE RECURSIVE RELATIONSHIP 3.1. Classification of the recursive relationship In a recursive relationship, a set of instances can take on a unique role of two different roles in the same relationship [7]. Checking the role allows us to classify all of the recursive relationships: symmetric or asymmetric. A recursive relationship is symmetric when all of instance participating in the relationship has a unique role and has the same semantics [8]. Accordingly, if R is a recursive relationship which is symmetric then role1 ≡ role2, in which role1 and role2 are the name of two roles of the recursive relationship R. Conversely, if R is not symmetric, we call this relationship is asymmetric. For example, the recursive relationship such as dancePartnerOf, spouseOf, siblingOf and isFriendOf are the symmetric relationships. Formally, we can define a symmetric recursive relationship as follows: There is the re- cursive relationship R on the entity type E, then R is called the symmetric if: ∀ e1, e2 ∈ E : if (e1, e2) ∈ R then (e2, e1) ∈ R. Conversely, R is called the asymmetric, if: ∃ e1, e2 ∈ E : (e1, e2) ∈ R and (e2, e1) /∈ R. 3.2. Transformation of the symmetric recursive relationship without attributes With a symmetric recursive relationship without attributes, because there is only one role, transforming into OWL similar by as transforming the binary relationship, but only add an object property. Rule EER12. For the symmetric recursive relationship R without attributes on the entity E with role Role, we have transformation rules as follows: - Add the object property with its name is the name of role in the relationship, domain and range is class C (E ), set the symmetric characteristics and other possible characteristics may have of the relationship; - With the relationship 1:1, add the maximum cardinality restriction by 1; - With the minimum cardinality constraint is 1-1, add minimum cardinality restriction by 1 to this object property. As in Figure 10, a person may be able to choose whether to pay taxes together or not. This relationship is symmetrical because there will be two cases: - Two people e1, e2 are joined in the relationship: e1 pay the same as e2 and e2 pay the same as e1. - Two people e1, e2 are not joined in the relationship: e1, e2 do not pay taxes with anyone. Apply these transformation rules of the symmetric recursive relationship 1:1 TaxJointly- With of the entity type Person as in Figure 10. 3.3. Transformation of the symmetric recursive relationship with attributes With symmetric recursive relationship with attribute, consider the relationship as an entity, so when transforming into OWL similarly to transforming the binary relationship with attribute, by adding new class and two inverse object properties. TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 87 Figure 10. An example of transformation of the symmetric recursive relationship without attributes Rule EER13. Let us consider the symmetric recursive relationship R with attribute attR of the entity type, we have the transformation rules as follows: - Add the class C (R), the attributes attR of the relationship is transformed into the datatype properties attR of class C (R); - Add two inverse object properties EHasR, ROfE representing the relationship between class C (R) and class C (E ). Two properties are reflexive and other characteristics of the relationship R. The object properties ROfE has min and max restriction is 1; - Set min/max restriction on the object properties EHasR corresponding with value on role of the min/max cardinality constraint which different from 0 and N . If R is the binary relationship N : N , add the object properties ROfE to set of key properties of class C (R). 3.4. Transformation of asymmetric recursive relationship without attributes Rule EER14. Considering the symmetric recursive relationship R without any entity on the entity type E with two roles are role1 and role2, we have the transformation rules as follows: - Add a pair of inverse object properties with its name are role1 and role2, domain and range is the class E ; which has the characteristics of corresponding relationship; - Each role with min/max cardinality constraint different from 0 and N , add the min/max cardinality restriction corresponding to the object properties. Considering the entities Person is the set of married people and the relationship mar- riedTo. One of the married people is married to only one person. That is, e1 married e2, in contrast e2 married e1. Two people e1, e2 are not married to any other person. Apply these transformation rules of the asymmetric recursive relationship without attribute marriedTo of the entity type Person as in Figure 11. 3.5. Transformation of the asymmetric recursive relationship with attributes With an asymmetric recursive relationship with attributes, in consideration of the rela- tionship as an entity, in order to ensure semantics during transformation, it is necessary to 88 VO HOANG LIEN MINH, QUANG HOANG Figure 11. An example of transformation of the asymmetric recursive relationship without attributes transform the asymmetric recursive relationship into a new class, and the attribute of the asymmetric recursive relationship that transforms into the datatype property of the newly class. Because the relationship has two different roles, it is necessary to create two pairs of two inverse object properties which represent the two roles of the relationship. Rule EER15. Considering the asymmetric recursive relationship R with attribute attR on the entity type E with two roles role1 and role2, we have transformation rules as follows: - Add the class C (R), attributes attR of the relationship R is transformed into the datatype properties attR of class C (R); - Add two inverse object properties Erole1R, Rrole1E representing a relationship between class C (R) and C (E ) with roles role1. Add two inverse object properties Erole2R, Rrole2E representing relationship between class C (R) and C (E ) with roles role2. Set min/max cardinality restriction for two properties Rrole1E and Rrole2E is 1; - Each min/max cardinality different 0 and N on role role1 of the relationship R, add min/max cardinality restriction corresponding to the object properties Erole1R; - Each min/max cardinality different 0 and N on role role2 of the relationship R, add min/max cardinality restriction corresponding to the object properties Erole2R; - If the relationship R is N : N , add two object properties Rrole1E, Rrole2E to the set of key properties of class C (R). Consider the relationship guards on the entity type Person. One can guard one person or no one, whereas one can be guarded by one person or without guard. This relationship contains two roles: guard and isGuardedBy. Apply these transformation rules of the asym- metric recursive relationship with attribute guards of the entity type Person as in Figure 12. 4. TRANSFORMATION OF TimeER INTO OWL ONTOLOGY TimeER model is an extension of the ER model; the construction of the ontology to represent for the temporal aspects from TimeER model is the problem inspiring many re- searchers. Thus, the transformation TimeER model into OWL ontology will be as follows: first, transform the components without temporal aspects on TimeER model (including the temporal entity type) into OWL ontology. After that, create the OWL ontology to represent for the temporal aspects in TimeER model. The purpose of this step is to represent the TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 89 Figure 12. An example of transformation of the asymmetric recursive relationship with attributes data and temporal aspects constraint on TimeER model. Finally, transform the temporal components on TimeER model into OWL ontology. 4.1. Initially ontology for representing the temporal aspect In OWL, no component has the same meaning as the temporal components in the Ti- meER model. To support the presentation of temporal aspects in OWL, we need to create an additional class InstantDateTime and object properties that represent the relationship between the class owl:Thing class and the class InstantDateTime. Create class InstantDateTime representing for a timeline. In this class, create the functi- onal datatype property hasDateTime with minimum cardinality restriction set to 1. This propertys range is xsd:dateTime and it is the key properties of a class InstantDateTime. Create six functional object properties with minimum cardinality restriction set to one: hasVTs, hasVTe, hasLSs, hasLSe, hasTTs, hasTTe; range set to class InstantDateTime and domain set to the class owl:Thing. Six properties represent the relationships between class owl:Thing and class InstantDateTime. 4.2. Transformation of the temporal aspect on TimeER model 4.2.1. Transformation of temporal entity type To represent the temporal aspects of an entities in the TimeER model, we can add a class, two inverse object properties, and some generated attributes. Rule TimeER1: Each temporal aspect XX of the entity type E : - Add the class C (E XX ); - Add two inverse object properties: EHasXX with domain is class C (E ) and range is class C(E XX); XXOfE with domain is class C(E XX) and range is class C(E), simultaneously XXOfE has functional characteristics and minimum cardinality restriction is 1; 90 VO HOANG LIEN MINH, QUANG HOANG Figure 13. Ontology represents the temporal aspects - Set of key properties of class C(E XX) include properties XXOfE and some properties for representing the temporal restriction which depends on temporal aspect XX as in Table 2. Proof : 1. According to [4], relating to the transformation of the temporal entity type of TimeER model into ER model, the entity type E with temporal aspect XX is transformed into the non-temporal entity type E which has a non-temporal attributes XX. Simultane- ously, create the weak entity type W (E XX ) which has the identifying relationship with the entity E, the purpose is to represent for the support temporal aspect XX for each entity type E. The partial key of the weak entity type W (E XX ) includes some attributes for representing the temporal constraint, depending on type of the temporal aspect XX. 2. According rule EER8, the entity type E is transformed into class C (E ), the attributes of the entity type E are transformed into the datatype properties of class C (E ). The weak entity type E XX is transformed into class C (E XX ), add two inverse object properties EhasXX and XXOfE representing for the relationship between class C (E ) and class C (E XX ) with domain and range as shown in Table 1. From (1) and (2), it is proved.  To manage the time that an entity exists in reality and the time that the entity/event is in the database, so the temporal aspect of the entity type Employee was transformed to class Employee LT and two inverse object properties shown in Figure 14. We add the class Employee LT and two inverse object properties which describe the relationship between the two classes. In particular, the min cardinality of the object property TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 91 Figure 14. Transformation of the temporal aspect of the entity type Table 2. The key attributes corresponding with the temporal aspect Temporal aspect The key attributes VT hasVTs LS hasLSs TT hasTTs LT hasLSs, hasLSe, hasTTs BT hasVTs, hasVTe, hasTTs LTOfEmployee is set to 1 to ensure that an individual in the class Employee is associated with an individual in the class Employee LT. 4.2.2. Transformation of the temporal attributes of the entity type In the case of temporal attribute of an entity, in consideration of the attribute as an entity, during the transformation into OWL, we add new class and two inverse object properties. Rule TimeER2. Each attribute attA has the temporal aspect XX of the entity type E : - Create class C (attA XX ), transform attribute attA to the datatype properties of class C (attA XX ); - Create two inverse object properties: attAHasXX with domain is class C (E ) and range is class C (attA XX ); XXOfattA with domain is class C (attA XX ) and range is class C (E ). XXOfattA has functional characteristics and minimum cardinality restriction is 1; - Set of the key properties of class C (attA XX ) includes property XXOfattA and some properties which represent the temporal restriction depend on the temporal aspect XX as shown in Table 2. Proof : 3. Accroding [4], relating to the transformation of the temporal entity type of TimeER model into ER model, the attribute attA with temporal aspect XX is transformed into the weak entity type W (attA XX ) which has the identifying relationship with the entity E, the purpose is to represent the temporal aspect XX of attribute attA for 92 VO HOANG LIEN MINH, QUANG HOANG each the entity type E. The weak entity type W (attA XX) have the attributes attA, partial key and some attributes for representing the temporal constraint, depending on the type of the temporal aspect XX. 4. According to rule EER8, the entity type E is transformed into class C(E). The weak W (attA XX) is transformed into class C(attA XX), add two inverse object properties attAHasXX and XXOfattA which represent the relationship between class C(E) and class C(attA XX), their domain and range as shown in Table 1. From (3) and (4), it is proved.  Figure 15. Transformation of the temporal attributes of the entity type In Figure 15, the attribute Salary with temporal aspect BT will be transformed to class Salary BT and two inverse object properties, the attribute Salary will be transformed to the datatype properties of class Salary BT. 4.2.3. Transformation of the temporal relationship In the case of transforming of a temporal relationship, the relationship is considered as an entity. To represent a temporal relationship in the TimeER model, we can add new class, two inverse object properties, and some generated attributes. Rule TimeER3. Each relationship R has the temporal aspect XX among the entities Ei: - Create class C (R); - Corresponding to each entity type Ei in the relationship R, create two inverse datatype properties for representing the relationship between class C (R) and class C (Ei): EiHasR with domain is class C (Ei) and range is class C (R); ROfEi with domain is class C (R), range is class C (Ei), set functional characteristics and minimum cardinality restriction is 1. If min/max cardinality constraint of the entity Ei different from 0 and N then add the min/max cardinality restriction corresponding to property EiHasR; - If R is the 1-1 binary or recursive relationship then key of class C (R) consists of one of the two newly added whose range is the class corresponding with the entity in relationship R and some properties representing for the temporal restriction depend on the temporal aspects XX as shown in Table 2. TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 93 Proof : Because the relationship R is seen as an entity type E (R) then the temporal relati- onship becomes the temporal entity type. Therefore, apply Rules TimeER2, Rule TimeER3 is generated. Then, corresponding with instance of the relationship between entity e ∈ E1 and f ∈ E2 is an entity of the entity type E (R). Figure 16. Transformation of the temporal relationship In the relationship Worrk for with temporal aspect VT to manage a valid time to record the time that an event is true in reality, we transform to class WorkFor, and add two pair inverse object properties show in Figure 16.  4.2.4. Transformation of the temporal attribute of the relationship Rules TimeER4. Each attribute attR has the temporal aspect XX of the relationship R: - Add class C (attR XX ), the attribute attR is transformed into the datatype property of class C (attR XX ) following the rules EER2, EER3 and EER4; - Add two inverse object properties that represent the relationship between class C (R) and C (attR XX ): attRHasXX with domain is class C (R) and range is class C (attR XX ); XXOfattR with domain is class C (attR XX ), range is class C (R), and has functional cha- racteristics. Minimum cardinality restriction of two properties set to one; - Key of class C (attR XX ) consists of the properties XXOfattR and some properties which represent the temporal restrictions depend on the temporal aspect XX as shown in Table 2. Proof : Because the relationship E is seen as an entity type E (R), so the temporal attribute of the relationship will be the temporal attribute of the entity type E (R). Therefore, apply Rule TimeER2, Rule TimeER4 is generated. In the relationship in Figure 17, the manager takes over the management positions in the project. Each manager is in a different position. Each person can take many different positions in the projects they participate in. So we transform the temporal aspect VT of the attribute Position to class Position VT, and the attribute Position was transformed into the datatype property of class Position VT.  94 VO HOANG LIEN MINH, QUANG HOANG Figure 17. Transformation of the temporal attribute of the relationship 5. REVIEW Based on the set of rules presented above, we can easily transform any EER model to OWL ontology. It can be seen that in order to transform the entity, the rules EER1, and EER5 - EER7 rules should be applied; and for the properties transformation, the rules EER2-EER4 are applied; for the relationship transformation, the rules EER9 - EER15 are applied. Suppose that n is the number of entities type, l is the number of attributes of each entity type, m is the number of relationships in the model. We see that if the above- mentioned transformation entity rules are applied, the complexity of algorithm will be O(n), the complexity of the attribute transformation will be O(n × l) and the complexity of the transformation of the relationship will be O(m). Since the complexity of each rule is O(1), the complexity of the transformation method will be O(max (m, n × l)). 6. CONCLUSIONS This paper has proposed an approach to build the OWL ontology from EER model by using the rules to transform EER and TimeER model into OWL ontology. Particularly, this research focuses on clarifying, and categorizing semantics of the recursive relationships on the ER model, thereby put forward the rules for transformation of recursive relationship. Furthermore, the rules for transformation of TimeER model into OWL ontology, such as the temporal entity type, the temporal attributes of the entity type, the temporal relationship, and the temporal attribute of the relationship are also presented clearly. The transformation rules by the virtual model is presented, and the OWL ontology results is verified on Protg. TRANSFORMING EXTENDED ENTITY-RELATIONSHIP MODEL 95 Figure 18. An example of OWL ontology transform from Figure 1 96 VO HOANG LIEN MINH, QUANG HOANG REFERENCES [1] M. Fahad, “ER2OWL: Generating OWL ontology from ER diagram”, IFIP The International Federation for Information Processing, Beijing, China, October 19-22, 2008. [2] Sujatha R Upadhyaya, P Sreenivasa Kumar, “ERONTO: A tool for extracting ontologies from extended E/R diagrams”, ACM Symposium on Applied Computing, Santa Fe, New Mexico, March 13 - 17, 2005, pp. 666-670. [3] Igor Myroshnichenko, M.S., Marguerite C. Murphy, Ph.D., “Mapping ER schemas to OWL ontologies”, Semantic Computing. ICSC ’09. IEEE International Conference, Berkeley, CA, USA, Sept. 14-16, 2009, pp. 324-329. [4] Q. Hoang and P. Thuong, “Expressing a temporal entity-relationship model as a traditional entity-relationship model”, in 7th International Conference, ICCCI 2015, Madrid, Spain, Sep- tember 21-23, 2015, pp 483–491. [5] Toan Van Nguyen, Hoang Lien Minh Vo, Quang Hoang, and Hanh Huu Hoang, “A new method for transforming TimeER model-based specification into OWL ontology”, in 8th Asian Confe- rence on Intelligent Information and Database Systems ACIIDS, Da Nang, Viet Nam, February 27, 2016, volume 642, pp 111-121. [6] Q. Hoang, H.L.M. Vo, and V.H. Vo, “Mapping of nested multi-valued composite attributes: an addition to conceptual design for XML schemas”, in Asian Conference on Information Systems 2014 - ACIS 2014, Nha Trang, Viet Nam, 2014. [7] James Dullea, Il-Yeol Song, and Ioanna Lamprou, “An analysis of structural validity in entity- relationship modeling”, Data and Knowledge Engineering - DKE, Netherlands, vol. 47, no. 2, pp. 167–205, 2003. [8] Grigoris Antoniou, Frank van Harmelen, “Web Ontology Language: OWL”, in Handbook on Ontologies, Springer, 2009, pp. 91–110. [9] M.K. Smith, C. Welty, and D.L. McGuinness, “OWL web ontology language guide”, W3C Re- commendation, 11-12/ 2004. [Online]. Available: https://www.w3.org/TR/owl-guide/. [Accessed 2 2 2017]. [10] B. Motik, P.F. Patel-Schneider, B. Paria, C. Bock, A. Fokoue, P. Haase, R. Hoekstra, I. Horrocks, A. Ruttenberg, U. Sattler , et al., “OWL 2 web ontology language: structural specification and functional-style syntax”, 11 -12/2012. [Online]. Available: https://www.w3.org/TR/owl2- syntax/. [Accessed 8 2 2017]. [11] H.C. Cheng, J.L. Cheng, and T.L. Chin, “Nonlinear system control using adaptive neural fuzzy networks based on a modified differential evolution”, Systems, Man, and Cybernetics, Part C: Applications and Reviews, Transactions on IEEE, vol. 39, no. 4, pp. 459-473, 2009. [12] Pasapitch Chujai, Nittaya Kerdprasop, Kittisak Kerdprasop, “On transforming the ER model to ontology using protg OWL tool”, International Journal of Computer Theory and Engineering, vol. 6, pp 484-489, 2014. [13] Hoang Huu Hanh, Le Manh Thanh, Course of Semantic Web, Vietnam Education Publishing House Limited Company, 2012 (in Vietnamese). [14] R. Elmasri, S.B. Navathe, Fundamentals of Database Systems, Addison-Wesley, 7th edn, 2015. [15] G. Antoniou, and F.V. Harmelen, A Semantic Web Primer, Cambridge: The MIT Press, 2004. Received on June 18, 2017 Revised on June 01, 2018