Recently XML is the standard format used for the exchange of data between information systems and is also frequently applied as a logical database model. If we use XML as a logical database model we need a conceptual model for the description of its semantics. However, XML as a logical database model has some special characteristics which makes existing conceptual models as E-R or UML unsuitable. I n this paper, the current approaches to the conceptual modeling of XML data are described in an uniform style. A list of requirements for XML conceptual models is presented and described approaches are compared on the base of the requirements.
Today XML is used for the exchange of data between information systems and
it is frequently used as a logical database model for storing data into databases.
If we use XML as a logical database model we need a conceptual model for
modeling XML data. There is the Entity-Relationship (E-R) [
These features cannot be properly modeled in the E-R model. T here are
some approaches, for example Extended E-R [
Suppose an E-R diagram with a relationship type Enroll between two entity types Student and Course representing courses enrolled by students. Each student may enroll zero or more courses and each course may be enrolled by zero or more students. The diagram is shown in Figure 1(a). Figures 1(b), (c), and (d) show possible representations of the relationship type in a hierarchical structure. Oriented arrows denote a nesting. There is not the best nesting of the concepts. The nesting of courses into students illustrated by Figure 1(b) is suitable when we need to see students and the courses they enrolled. The nesting of students into courses illustrated by Figure 1(c) is suitable when we need to see courses and the students enrolled in them.
The previous example shows another difference between the conceptual level of XML and the E-R model. This difference is not in the structur e but it is in the usage of XML. It is shown that there may be many ways of how to use entity types connected together by a relationship type. If we represent data in the form of XML, each of these ways may require another hierarchical ordering of the entities. However, this feature cannot be effectively modeled by the E-R model.
Another possibility of how to model XML data is to start from a hierarchical
structure. This approach may be called the hierarchical approach. There are
conceptual models based on the hierarchical approach, for example X-Entity
[
The hierarchical approach is able to solve the mentioned problem with different views of the same data. For each of the views there is a separate tree. However, a problem with the modeling of attributes of relationship types or with the modeling of n-ary relationship types, effectivelly solved in the E-R mode l, arises. Another problem arises when deciding which of hierarchical organizations of the same data is the best to select as the basic organization used for the data storage.
The goal of this paper is to describe the existing conceptual models for XML
based on the E-R model and on the hierarchical approach. Ther e are approaches
based on the UML (Unified Modeling Language) [
Section 2 introduces the list of requirements for conceptual models for XML. Section 3 describes representatives of the conceptual models for XML based on the well-known E-R model. Section 4 describes representati ves of the hierarchical conceptual models for XML. Section 5 compares the described conceptual models against the requirements introduced in Section 2. 2
Requirements for Conceptual Models for XML Requirements for conceptual models for XML are summarized in this section. There are two groups of the requirements described. The first group consists of general requirements covering general goals of the XML conceptual modeling. The second group consists of modeling constructs requirements covering requirements on what kinds of modeling constructs should XML conceptual models support. 2.1
be independent on a certain XML schema language (XML Schema [
Formal foundations The modeling constructs of the conceptual model should be described formally, which allows to compare the model with other conceptual models or to describe the operations on the model structures and modeled data (for example, data transformation between two conceptual schemata or their integration).
Graphical notation A user-friendly graphical notation for the formal modeling constructs should be offered by the conceptual model.
Logical level mapping There should be algorithms for mapping of the
conceptual modeling constructs to the XML logical level. The logical schema
should implement as many integrity constraints arised from the conceptual
schema as possible. It may require the usage of more than one XML schema
language for the logical level description (XML Schema and Schematron [
Different structures on the logical level The XML logical level is
hierarchical. However, there are different users with different requirements
accessing the modeled data on the logical level. Hence, there can be different
hierarchical views of the same data. Each of the views suits to different
requirements. It should be possible to model the different hierarchical views
on the conceptual level and translate them to the corresponding views on
the logical level. Moreover, there should be algorithms allowing automatic
translation of data from one logical view to another logical view (using XSLT
[
Semantic web mapping With the increasing usage of the semantic web
technologies the problem of publishing data in the form of RDF [
Hierarchical structure Although it can be useful to keep a document designer out of the hierarchical structure of XML data on the conceptual level, the conceptual model should offer modeling constructs for modeling nesting explicitly. For example, aggregation relationship types can be used. However, non-hierarchical relationship types (for example, a ssociation relationship types or references) should be offered too. The conceptual model should introduce contructs for modeling a recursive structure.
Cardinality for all participants The hierarchical structure of XML data restricts the specification of cardinality constraints only to the nested participants of the relationship type. However, it should be possible to specify cardinality constraints for the all participants on the conceptual level. N -ary relationship types For the same reason, the modeling of n-ary relationship types and their translation to the XML logical level is problematic. However, it should be possible to model n-ary relationship types on the conceptual level.
Attributes of relationship types For the same reason again, the modeling of attributes of relationship types is problematic. Nor the nesting nor the concept of referential integrity on the XML logical level do not allow to directly express attributes of relationship types. However, the conceptual model should allow to model attributes of relationship types.
Ordering XML is ordered and this property should be propagated to the conceptual level. It should be possible to express the ordering on values of attributes, the ordering on concepts connected with another concept (for example, a book has a title page first, followed by an abstract, chapters, appendixes and a bibliography in this order) and the ordering on a participant of a relationship type (for example, the list of authors of a book or the list of chapters of a book are ordered).
heterogeneous structure. The conceptual model should introduce constructs for modeling such a structure. For example, variant-valued constructors for constructing attributes or disjunctive relationship types should be introduced.
Document-centric data The difference between the conceptual models for
XML and the other conceptual models is that the conceptual models for
XML must allow to model document-centric data. It means that not only
the real-world objects with attributes and relationships b ut also the certain
parts of documents are modeled on the conceptual level. Hence, there should
be corresponding modeling constructs offered by the conceptual model. It
means to allow attributes and relationships of a given concept to be mixed
with a text when represented in a document content. However, the mixed
content should not be restricted as it is restricted by XML Schema. Some
form of generalized mixed content should be introduced allowing to specify
where the text values may appear exactly (as it is possible in Relax NG [
Reuse of content The reuse of content should be supported by the conceptual model. For example, the concept inheritance (modeled by IS-A relationship types in E-R, for example) supports the reuse of con tent. However, the conceptual model may be inspired in the XML Schema language and may support named types and named groups of concepts on the conceptual level.
Integration of conceptual schemata XML data are often used for the data integration. However, it can not be done effectivelly and automatically without the support on the conceptual level. A conceptual model for XML should offer modeling constructs to support an integration of schemata on the conceptual level and it should allow to merge different conceptual schemata to an overall conceptual schema. Further, it would be useful to generate XSLT transformation scripts to translate data corresponding to one conceptual schema to data corresponding to another conceptual schema. 3
ER- Based Conceptual Models for XML
In this section, we describe two representatives of the conceptual models for
XML based on the E-R model. The first representative is the Ext ended E-R
model [
Extended E-R model proposed by Badia in [
In a graphical representation, an optional attribute is connected to the corresponding entity type by a solid line with two dashes crossing it. A choice of attributes is expressed by marking the choice with an upward triangle, with the choices in the opposed side of the triangle.
Figure 2 displays the entity type Student having the optional attribute
phone and the exclusive choice attribute involving the attributes hosteladdr
and homeaddr, i.e. each student has a hostel address or a home address but not
both.
EReX is an extension to the E-R model proposed by Mani in [
The extending modeling constructs of the EReX model are demonstrated by the schema in Figure 3. It displays the categorized entity type P erson and its categories Student and P rof essor. The key of P erson is empty. Further, there are the entity types Book and P aper connected with P rof essor by the relationship types AuthorOf B and AuthorOf P , respectively. Attributes of Book and P aper are not displayed. There is an ordering specified on the entity types Book and P aper in the relationship types AuthorOf B and AuthorOf P , respectively. It means, that the authors of a given paper or a given book are ordered.
The total coverage constraint Student + P rof essor = P erson specifies, that each person is a student or a professor and there are no other persons. The exclusive coverage constraint Student| P rof essor specifies that students and professors are disjoint. The total coverage constraint AuthorOf B.pbook + AuthorOf P.ppaper = P rof essor specifies that each professor is an author of some paper or book.
Hierarchical Conceptual Models for XML The extensions of the E-R model allow to model conceptual sch emata with a graph structure. However, XML schema languages allow to express relationship types only by nesting and references. It is possible to express all the relationship types from an E-R schema by references, but it leads to flat sch emata and the advantages of the hierarchical structure of XML are not utilized. On the other hand, if the hierarchical structure is used to express relationship types in a conceptual schema the problem with the decision about what to nest arises. Another problem is how to represent n-ary relationship types and attributes of relationship types.
In this section we describe a basic hierarchical conceptual model for XML
first. In the next subsections, we describe two representatives of the conceptual
models for XML based on the hierarchical approach. The first representative is
the ORA-SS model [
The basic hierarchical conceptual model for XML can be easily defined as a restriction of the E-R model where only the binary relations hip types with cardinality types (1, 1) : 1 or (1, 1) : N and without attributes are allowed. Each relationship type is oriented from the entity type with the arbitrary cardinality called the parent participant to the entity type with the cardinality (1, 1) in the relationship type called the child participant. We say that the child participant is nested in the parent participant. This kind of relationship types may be called nesting binary relationship types. When modeling XML data, the nesting binary relationship types are represented by a nesting of elements on the XML logical level. They express a hierarchical structure on the XML logical level explicitly on the conceptual level. However, the semantics of nesting relationship types do not have to be only ”part-of”. It may be a general association too.
Such restrictions are too strong and do not allow to model conceptual schemata with richer semantics. Nor n-ary relationship types, nor attributes of relationship types can be modeled. Moreover, lots of redundancies may appear in schemata. There are some approaches extending this basic hierarchical model described in the following subsections. 4.2
ORA-SS
ORA-SS is a rich hierarchical conceptual model for XML propo sed by Dobbie
et al. in [
It allows to model irregular structure.
Figure 4 displays an ORA-SS schema representing professors as employees of departments and professors as members of projects. Each professor is employed by exactly one department and each department employes one or more professors. Each professor is a member of zero or more projects and each project has one or more members. There is a ternary relationship type AuthorOf between the object types P roject, P rof essor and P aper. It represents papers written by a professor participating in a project. Each professor is an author of zero or more papers in a project and each paper has one or more authors. For each project there is a list of member professors and for each member professor there is a list of papers he wrote during his work in the project. Moreover, there is the attribute pages of the relationship type AuthorOf . For a professor being an author of a paper in a project the value of pages is the number of pages the professor wrote in the paper. However, the attribute pages cannot be directly assigned to the relationship type AuthorOf . It must be assigned to the nested object type P aper. For each P rof essor instance nested in a P roject instance there must be a P rof essor instance nested in a Department instance containing the name and address values of the professor. This is modeled by the reference between the object types.
The semantic network model for XML was introduced by Feng et al. in [
Only binary hierarchical relationship types without attributes can be modeled in the semantic network model. Moreover the parent participant cardinality constraint of a hierarchical relationship type must be equal to (1, 1). Beside the hierarchical relationship types it is possible to use non-h ierarchical relationship types for modeling associations.
Different constraints can be specified in the semantic network schema for XML. Constraints can be specified over a node, over an edge and over a set of edges. These constraints are a uniqueness, order, disjunction, etc.
Figure 5 displays a semantic network schema. There are departments represented by the node Department and professors in the departments represented by the node P rof essor. The content of the node P rof essor is ordered. For each professor there are the papers he wrote represented by the node P aper. Each paper may be composed of chapters or sections, but not both (the exclusive constraint). The courses offered by a department are represented by the node Course. P rof essor is associated with Course. It represents the relationships between a professor and the courses he teaches.
Comparison of Described Conceptual Models In this section, we compare the conceptual models mentioned in this paper. The comparison is made against the general requirements and the modeling constructs requirements introduced at the beginning of the paper.
There are two comparative tables. Table 1 compares the models against the general requirements and Table 2 compares the models against the modeling constructs requirements. The well-known E-R model and the b asic hierarchical model are compared too.
We are not able to decide, which of the previous two approaches (E-R extensions, hierarchical modeling) is better for the conceptual modeling of XML data. Conceptual models based on the E-R model allow user to c reate a schema with no metadata redundancy, but there is the problem with the modeling of the specific XML features. Hierarchical conceptual models solve the problem with a hierarchical structure of XML, but there arises problems such as data and metadata redundancy, modeling of attributes of relationship types, and modeling of n-ary relationship types.
There are requirements that are not met by the described models. The modeling of document centric data and the reuse of content is problematic. The important requirement on the integration of conceptual schemata is solved only by the ORA-SS model. None of the models solves the problem of t he integration with the semantic web technologies. − − − − − − 1 formal description is missing − − − − − − √ − − − − √ √ √ − − √ − − − − − − − − − − on the content of a concept − − − − √ 2 − √ −
− on the participant of a relationship type − − √ √ − √ 3 − − · irregular and heterogeneous structure √ −1 variant-valued attribute constructor √
√ − − − − − − disjunctive constraints on relationship types − √ 4 √ 4 − − √ − − · document-centric data − basic mixed content
√ − − − − − − −
− generalized mixed content − − − − − − − · reuse of content
− IS-A or the category concept √ √ √ √ √ √
− − named types and groups of concepts − − − − − − − · integration of conceptual schemata − modeling constructs
√ − − − − − − algorithms for merging schemata − − √ − − − − − − − − − − √ − − − − − −
− − − − − − − − − algorithms for the data translation between schemata − − − − − − − − − 1 with the complex attributes extension 2 unordered content is restricted by the restrictions of xsd:all 3 only by ordered attributes, native XML ordering is not utilized 4 using the category concept 5 indirect modeling using hierarchical relationship types is possible √ √ − 5 − 5 √ √ √ √ √ − − √ − √ √ √ − √ − − √ √ − √ √ − − √ − − − −
In this paper, we describe a state of the art of the conceptual modeling for XML. There has been several papers proposing new conceptual models for XML. We selected some representants of the models and describe them in this paper. We propose a list of requirements conceptual models for XML should satisfy and compare the mentioned conceptual models against the requirements.
The comparison of the models shows that there is a poor support for some specific XML features as ordering or mixed content as described by the modeling constructs requirements proposed in Section 2. Moreover, the models poorly concentrates on the usage of conceptual schemata for the data integration and the integration with the semantic web technologies as described by the general requirements proposed in Section 2.
For these reasons, there is an open space for a research in the area of the conceptual modeling for XML. Not only new modeling constructs should be proposed to support specific XML features. The utilization of the conceptual models for the data integration between different data sources including semantic web resources should be explored too.