This article provides an overview of the data transformation solutions needed to integrate various models and data schemas. An example of mutual mapping of ontologies on relational databases is considered. The scientific basis of the presented data models is described. Also set goals and objectives for data integration. The article describes the problems and further steps to explore the possibility of data transformation in computer-aided design systems. The main problem of transforming a relational database into an ontology is the lack of metadata. With the inverse transform, some data and metadata may be lost. In addition, there are several approaches to the transformation of ontologies into relational databases. OWL is used as the ontology description language. The article is useful to architects and software designers, as well as knowledge engineering specialists.
The present stage of human development is characterized by an eno rmous speed of development. At the same time, science, originally formed as a practical application of thought, has long been theoretical and is increasingly divorced from practice, and therefore also technology. However, no technological task is possible without a scientific basis.
In particular, such a merger and separation of science and technology is observed in the description of our world, namely in the representations of knowledge and data.
The most famous data access technology, invented in the 20th century, is the relational database and the language for managing and manipulating SQL. This technology is widely used due to high speed, reliability and unification of access. Unlike many modern programming languages, the syntax of SQL, firstly, does not indicate specific commands for obtaining data and, secondly, is practically independent of the specific implementation. It is also important to note that the speed of data manipulation was achieved thanks to the mathematical apparatus of relational algebra or relational calculus. Many modern non-relational DBMSs (also called NoSQL) do not have the same coherent formal description, therefore, they are not as widely used as SQL.
Another approach to data presentation are ontologies with a scientific basis in the form of ontological analysis and descriptive logic. It was thanks to the descriptive logic that it was possible to express in a formal form, and then implement all sorts of dictionaries, as well as identify contradictions in the concepts of ontologies through reasoners. The practical application of ontologies is the formation of thesauruses, as well as an attempt to present the meaning on the world wide web (semantic web). The last application is especially important in the world of the Internet of thin gs and search engines.
However, it is worth noting that although relational databases and ontologies allow knowledge to be presented with some degree of similarity, these approaches developed in parallel and do not find many points of contact precisely bec ause RDBMS is the development of technology, and ontologies are the development of science.
We should also consider the issue of the documentability of modern information systems. To summarize, the documentation is a representation of the work of the syste m in a form convenient for a person. That is, it is a linguistic mapping of data models and business processes described within the system. At the same time, an ontology or a relational database (the latter, however, with some limitations) can be used to d escribe the subject domain.
It turns out that there are at least three important and different descriptions of the domain: a relational representation (technological), an ontological representation (scientific), and documentation (linguistic). And all representations denote a single whole, but in different ways. Therefore, all data models must be consistent with each other, or, in other words, be expressed through each other or through some other knowledge representation model.
Chapter 2 analyzes the various schemas and data representations to select the main schema through which others can be presented. The third chapter deals with the problem of data transformation using the example of ontologies (OWL technology) and relational databases. The fourth chapter describes the theoretical foundations of data transformation. The conclusion describes the results of the analytical review and describes the subsequent course of research in the current field. 2
When there are several representations (schemas) of data, it is necessary to choose one through which all others can be expressed.
The concept of data integration between various schemes is described in [
Also, in this paper several data conversion processes are considered: LAV (local as view) and GAV (global as view).
The LAV approach is based on the idea that the concepts of each source scheme should be described within the global scheme. This is possible if the global scheme is stable and well represented. An ontology can act as a global scheme. Also, this approach makes it easy to expand the system by adding new sources. It is only necessary to describe their model in global terms.
The GAV approach, on the contrary, states that the concepts of a global scheme should be expressed in terms of sources. This is only possible if the sources are stable. In other words, mapping with this approach explicitly states how and where to get data in terms of a global scheme. This provides for a declarative approach, since when manipulating data, we do not explicitly indicate where to get them, but only refer to the global schema.
Both approaches can be used in building data integration between the schemas. In
addition, both approaches can be transformed if desired, which is also described in [
In our case, the LAV approach will be chosen, since it has a stable core - a global scheme. Further, ontology with justification in the form of description logic will be considered as such a scheme. 3
With the modern development of science and technology, there are many data models that can provide the same information about the real world in various forms.
The most interesting for the task of data transformation are relational database management systems and ontologies, since both models have a scientific basis: relational algebra or relational calculus in a RDBMS, and display logic in ontologies. 3.1
There are many strategies for transforming relational schemes into ontology. For
example, one can single out the concept of direct mapping from [
Another technology for transforming a relational scheme into metadata, also
developed by W3C, is the R2RML language [
There are also scientific papers describing the transformation of relational data
schemas into ontologies, and not only into RDF metadata. For example, a comprehensive
review of current developments in such a transformation is given in [
Based on [
The categories described above are highlighted to correctly define the transformation rules. For example, entities without foreign keys can be transformed into ordinary ontology classes.
It is also necessary to consider the limitations of drying. For example, the constraint
constraint (NOT NULL) is set using MinCardinality (in OWL notation), and the
uniqueness constraint (UNIQUE) is set using the InverseFunctionalProperty property
[
Some transformation algorithms from RDBMS to ontologies (for example, in [
When constructing matching rules, it is worth bearing in mind the correctness (in other words, relationality) of the design of the original data scheme. There must be foreign keys between entities, since without them, ontology will degenerate into a set of concepts without any connections. It is also important to give descriptions and literate restrictions (at least on compulsion and on uniqueness) in the tables, since wit hout this, it will be impossible to build inference systems in ontology. 3.2
The inverse approach to translating schemas is to translate from an ontology scheme into a relational data model.
The reviews [
The main problems here are either limitations of the approach (for example, support only of the basic constructs of the OWL language), or initial tuning of the ontology, or poor quality of the resulting relational data schema. 4
The basis of ontologies is descriptive logic. They represent a family of logical
formalisms based on the concepts of the concept (class of concepts), the role (relationship)
and the individual [
The scientific basis of relational databases is relational calculus or relational algebra.
According to Codd's theorem, query languages for relational data, relational algebra
and relational calculus are equally expressive [
However, in the case of comparison of description logic and relational algebra, it is
necessary to compile a terms comparison table. It is presented in table 1.
For example, let Person be the concept denoting a person, Female be the concept
“female being”, Male the concept “male being”. Also declared the role of hasChild. Then
the concept of Person ⊓ Female will designate a woman, and the concentrated Person
⊓ ∃ hasChild.Person will designate a person who has a child. Examples are taken fro m
[
Concept Role Individual Property M etadata Relation Foreign key Tuple Attribute Constraints Additionally, you will need to enter the PersonChild entity with the following attributes: person_id - person identifier; child_person_id is the ID of the person-child.
Thus, in order to obtain all women, it is necessary to apply the rela tional sampling operation (restrictions) with is_female = true. To select all parents, it is necessary to relation Person relational join operation with the PersonChild relation predicate Person.person_id = PersonChild.person_id .
Also, in this chapter, it is necessary to separately note the work [
At the current time, the transformation of data models is widely used in the construction of complex information systems. This is due to the incompatibility of different technologies with each other, as well as due to duplication of the description of the same subject area several times, for example, in the description of the relational data schema, in the ontology of the subject area, in artifacts of the information system(documentation).
To reduce the cost of maintaining various data models that describe the same subject area, it is necessary to correctly and reasonably convert data schemas.
At the current time, data transformation from both RDBMS to ontology (usually using OWL and the Protege editor) and reverse mapping are considered. However, the following tasks are poorly addressed: description and implementation of ontology transformation into a relational structure using DBMS metadata (for example, names of foreign keys, unique constraints, description of tables and fields); a description of the above transformations from a formal point of view; proof of the isomorphism of the descriptive logic and relational calculus