J. Software Engineering & Applications, 2010, 3, 629-643
doi:10.4236/jsea.2010.37073 Published Online July 2010 (http://www.SciRP.org/journal/jsea)
Copyright © 2010 SciRes. JSEA
Using an Ontology to Help Reason about the Information
Content of Data
Shuang Zhu1, Junkang Feng2
1,2Database Research Group, School of Computing, University of the West of Scotland, Paisley, UK; 2Business College, Beijing
Union University, Beijing, China.
Email: {shuang.zhu, junkang.Feng}@uws.ac.uk
Received March 25th, 2010; revised April 23rd, 2010; accepted April 25th, 2010.
We explore how an ontology may be used with a database to support reasoning about the “information content” of data
whereby to reveal hidden information that would otherwise not derivable by using conventional database query lan-
guages. Our basic ideas rest with “ontology” and the notions of “information content”. A public ontology, if available,
would be the best choice for reliable domain knowledge. To enable an ontology to work with a database would involve,
among others, certain mechanism thereby the two systems can form a coherent whole. This is achieved by means of the
notion of “information content inclusion relation”, IIR for short. We present what an IIR is, and how IIR can be
identified from both an ontology and a database, and then reasoning about them.
Keywords: Ontology, Information Content of Data
1. Introduction
Data mining techniques and tools are developed for
finding otherwise hidden knowledge from data, and little
seems to have been done on bringing “standard” domain
knowledge into such a process, which we envisage would
be helpful.
Ontologies as domain knowledge have been used in
many fields. We want to explore how an ontology may
help find hidden information from data. In this paper, the
focus is on how to link an ontology with a relation data-
base in order to reason about informational relationships
between data constructs in the database and those be-
tween domain objects captured by an ontology. This may
represent an innovative approach to knowledge discovery
in a database.
Ontology [1] as a term used in computer science was
started in the 1990’s. Compared with the development of
relational databases, it is a new scientific field. Ontology
offers an opportunity to give an open and standardized
description of database semantics with which we can
substantially improve the quality and utilization of data.
That is,
Ontology + Database = (Standards + Explicit Seman-
tics) + Database,
which leads to improved data utilization and data quality
Futhermore, semantic web [3] is a popular topic.
Through semantic web we attempt to provide users with
far better machine assistance than it is available now for
their queries. Semantically annotated web pages with
ontologies may assist reachers to achieve this purpose
Through our work, we obtain an ontology from the
DAML library, which represents some additional com-
mon knowledge, and link it with an existing database. In
terms of linking an ontology and a database though, in
the literature, we find a few different methods in using an
ontology to assist a query process.
It appears that one way to achieve this is that an on-
tology is invoked at the very beginning of a query proc-
ess [5] as shown in Figure 1. That is, it is through
re-writing a query in order to get more information. A
Figure 1. Invoking an ontology in the query processing
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
user query is translated into a set of queries with the help
of the ontology, which better fits the structure of the data
source. After query optimization strategies having been
applied on them, the resultant transformed queries are
equivalent to the submitted ones. Although seemingly a
promising approach, it is not concerned explicitly with the
information content of data, in which we are particularly
interested and wish to explore and make use of.
Another approach that we have investigated is where
an ontology is invoked in formulating a query process by
Munir et al. [6]. In their approach, firstly, an ontology is
generated based upon domain metadata including rela-
tionships between data in a relational database. Then
such an ontology is enriched with domain knowledge.
Secondly, ontology statements are translated into expres-
sions in the OWL-DL language. Thirdly, the expressions
are transformed into relational query statements. Finally,
map the domain ontology to a relational database (as
shown in Figure 2). Munir et al. [6] said little about the
mapping between the created ad hoc ontology and the
“standard” domain ontology if any, which we suspect is
done intuitively. This is however one of the topics in
which we are particularly interested.
We give an outline of our approach in Section 2. The
key notion is informational relationship and its formali-
sation IIR [7]. We describe in details how IIR may be
derived from a relational database and from an ontology
in Section 3, which make use of inherent and ad hoc
constraints between data constructs in a database and
between concepts in an ontology. We present a full ac-
count on how our ideas are tested by using some imple-
mentation in Oracle in Section 4. Finally we give con-
cluding remarks in Section 5.
2. Outline of our Approach
Our approach is to invoke an ontology when we work on
a database. Namely, when a user submits a query, we do
not change the query, but rather we involve the ontology
in the reasoning process per se that is required for an
swering the query (shown in Figure 3). Furthermore and
most importantly for us, the reasoning is carried out on
the basis of the notion of “information content” of data.
This notion is the work of Xu, Feng and Crowe in 2008
[8], which extends substantially Dretske’s [9] definition
of “information content” of a signal. In this paper, they
introduce another notion called IIR, as a formulation of
the notion of “information content” of data.
Xu et al. [8] define IIR as follows: “Let X and Y be an
event respectively, there exists an IIR, from X to Y, if
every possible particular of Y is in the information con-
tent of at least one particular of X”. Furthermore, they
define that “Let X be a event, the information content of
X, denoted I(X), is the set of events with each of which
X has an information content inclusion relation”. More-
over, they present a sound and complete set of inference
rules (IIR rules) for reasoning about information content
of data (states of affairs, or events in general). The six
inference rules are cited below.
1) Sum
If 12 n
YX XX then
IX Yfor i = 1, …, n
This rule says if it is the disjunction of a number of
events, then an event X is in the information content of
any of the latter. A trivial case is where X and Y above
are not distinct.
2) Product
If 12 ,
for i = 1, …, n then
This rule says that if an event X is the conjunction of a
number of events, then any of the latter is in the informa-
tion content of the former. A trivial case is where X and Y
above are not distinct.
3) Transitivity
This rule says that if the information content of an
event X includes another event Y, and the information
Figure 2. Ontology assisting the formulation of a query
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Figure 3. Ontology enhances reasoning about the information content in a database
content of Y includes yet another event Z, then the infor-
mation content of X includes Z.
4) Union
 
, then
This rule says that if the information content of an
event X includes another two events Y and Z respectively,
then the information content of X includes event YZ
that is the product of Y and Z. And it is in this sense that
Y and Z are in a “union”.
5) Augmentation
If 12 n
WW WW, Z is the product of a subset of
WW W then
This rule says that if 12 ,
WW W event Z is the
product of a subset of
,,, ,
WW W and the infor-
mation content of event X includes event Y, then the in-
formation content of the event WX formed by the
product of W and X includes the event ZY formed by
the product of Z and Y.
6) Decomposition
 
This rule says that if the information content of event
X includes event YZ that is the product of event Y and
event Z, then Y and Z, as separate events, are in the in-
formation content of X, respectively.
In this paper, we exploit the ideas above. That is, in a
way, we translate both the ontology and the database into
IIR and then reason about them as a whole. Put another
way, as what matters is information and IIR captures and
formulates it, so we look at both an ontology and a data-
base from the same perspective of IIR, and this enables
the two different things to work together. The overview
of our approach is illustrated in Figure 3.
On the very top of Figure 3, there is a block called
“information collection from the real world”. From this
information, knowledge about a domain of interest in-
cluding explicit business rules is arrived at. Domain
knowledge is then formulated as an ontology by using
software tools and languages.
Two different routes are there to deal with user queries.
If it is in a conventional query language then a query is
handled in a normal way. The dotted line indicates this
route. If that does not work, we would invoke the other
route, i.e., to invoke ontology and reasoning about IIR.
The second solution is the primary goal of this project,
which is indicated by the solid line arrow in Figure 3 of
“Customer query”“IIR closures”“Query results”.
The only difference between these two solutions lies in
the middle part of the procedure, on which we concen-
trate. Within the “Integration of IIR” section, there are
three different resources required to derive the “IIR”,
indicated by three arrows from “ontology”, “business
rule” and “database”, which are the origins of initial IIR.
Then there is a reasoning mechanism implemented in
PL/SQL of Oracle. The result of the reasoning is IIR
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
closures. Given an event A, the IIR closure of A, denoted
as A+, is the set of all events that are in the information
content of A, that is, if IIR(A, B), then B A+. IIR clo-
sures are the basis of answering queries in our approach.
Our work thus far shows that it is the additional rela-
tionships between data constructs especially “entities”
that are revealed and made available through using an
ontology that give us more and enlarged IIR closures
than those that would otherwise be based on the database
alone. This is how our approach makes a difference.
One of the main tasks is to derive IIR from the ontol-
ogy, the database and business rule, and then integrate
them as a whole. For instance, suppose that we have
IIR(A, B) (meaning the information content of A in-
cludes B) and IIR(C, D) from a relational database, and
IIR(E, F) and IIR(G, H) from an ontology. If we also
know that A and E are equivalent, then with Transitivity,
we get IIR(E, B) and IIR(A, F). Consequently A+ and E+
are enlarged.
We use Oracle [10] to implement this approach. An
ontology in OWL [11] can be translated into relational
tables [12]. Such tables do not hold data values however,
if the ontology is an unpopulated one. In such a case, the
involvement of an ontology results in additional objects
and additional relationships between objects that are rep-
resented by data in the original relational database. This
way, a query that does not have exact match with data in
the database may be answered. An ontology may add an
additional hierarchical structure to data in the database.
Furthermore, as said earlier, we use ontologies in a spe-
cial type of reasoning, i.e., reasoning about the informa-
tion content of data through a kind of special relationship
between data items and between data items and real
world objects, namely informational relationships, which
is captured and formulated as IIR between events (in
terms of probability theory). Thus, how to identify IIR
from an ontology becomes a key factor in our approach.
3. Deriving IIR
An IIR is a relationship between two states of affairs (i.e.,
events) such that one’s existence results in the certainty
that the other exists, and without the former, the latter is
not certain. Following Dretske 81 [9], we say that the
latter is in the “information content” of the former.
It would appear that to express IIR(X, Y) must be
based on and revolved around two elements. One is two
individual values (two individual parts or two sets of
groups) captured as X and Y, and the other is relation-
ships between X and Y.
We use part of a “university” database and part of on-
tology “Academic” to present how IIR can be derived
from a database and an ontology. Then the IIR are rea-
soned about by applying aforementioned Inference Rules.
The reasoning is implemented by a program.
3.1 Deriving IIR from an Ontology
According to characteristics of ontologies, these are two
different sources that may help the derivation of IIR. One
is concerned with relationships between “Classes” in an
ontology. The other is “ObjectProperty”.
3.1.1 IIR Derived from Classes
Generally, there are two different types of relationships
between classes from which IIR exist. One is “subClas-
sOf”, and the other “equivalentClass”. The syntax for
these two in an OWL ontology is as follows:
A relationship between “Class” and “subClassOf”,
<owl:Class rdf:ID=”Lecturer”>
<rdfs:subClassOf rdf:resource=”# Faculty”>
A relationship between “Class” and “equivalent-
<owl:Class rdf:ID=”Teachers”>
<owl:equivalentClass rdf:resource=”#Faculty”/>
The IIR could be derived from these two relationships
IIR(Class, subClassOf),
IIR(Class, equivalentClass) and IIR (equiva-
lentClass, Class).
As shown above, we have a relationship “Lecturer is a
subclass of Faculty, and Teachers is an equivalent class
to Faculty”. Hence we have IIR(Lecturer, Faculty),
IIR(Teacher, Faculty), IIR(Faculty, Teacher), and
IIR(Lecturer, Teacher).
3.1.2 IIR Derived from ObjectProperty
There are four different types of ObjectProperty rela-
tionships, which capture relationships between classes in
an OWL ontology. These are: “ObjectProperty”, “sub-
PropertyOf”, “equivalentProperty” and “inverseOf”.
As aforementioned, to create IIR needs two classes (X
and Y) from the ontology. As ObjectProperty represents
a relation for connecting two classes of “domain” and
“range” in an OWL ontology, an ObjectProperty already
contains a set of classes, which can be expressed as “Ob-
jectProperty=(‘domain’, ‘range’)”. Accordingly, we ob-
tain IIR(domain, range). That is, the IIR that can be de-
rived from these four types of ObjectProperty is all of the
IIR(domain, range)
Note that IIR must be of a many-to-one relationship
(including one-to-one). How to handle many-to-many is
to be addressed shortly.
The relevant syntax of OWL is as follows:
A relationship between “ObjectProperty”,
<owl:ObjectProperty rdf:ID=”research_by ”>
<rdfs:domain rdf:resource=”Professors”/>
<rdfs:range rdf:resource=”Projects”/>
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Then we have “IIR(domain, rang)”, for example,
IIR(Professors, Projects).
A relationship between “ObjectProperty” and
<owl:ObjectProperty rdf:ID=”research_in”>
<rdfs:domain rdf:resource=”Postgraduates”/>
<rdfs:range rdf:resource=”Projects”/>
<owl:ObjectProperty rdf:ID=”study_in”>
<rdfs:subPropertyOf rdf:resource=”research_in”/>
<rdfs:domain rdf:resource=”Postgraduates”/>
<rdfs:range rdf:resource=”Projects”/>
Then we have “IIR(domain, range)”, for example,
IIR(Postgraduates, Projects).
A relationship between “ObjectProperty” and
<owl:ObjectProperty rdf:ID=”attend_course”>
<rdfs:domain rdf:resource=”Student”/>
<rdfs:range rdf:resource=”Course”/>
<owl:ObjectProperty rdf:ID=”join_course ”>
<rdfs:domain rdf:resource=”Students”/>
<rdfs:range rdf:resource=”Course”/>
Then we have “IIR(domain, range)”’, for example,
IIR(Students, Course).
A relationship between “ObjectProperty” and “in-
<owl:ObjectProperty rdf:ID=”teache_of”>
<rdfs:domain rdf:resource=”Faculty”/>
<rdfs:range rdf:resource=”Course”/>
<owl:ObjectProperty rdf:ID=”instruct_by”>
<owl:inverseOf rdf:resource=”teaches_of”/>
<rdfs:domain rdf:resource=”Course”/>
<rdfs:range rdf:resource=”Faculty”/>
Then we have “IIR(domain, rang)”, for example,
IIR(Course, Faculty).
Moreover, there are relationships between classes and
“Not Null” FunctionalProperty, the syntax of which is as
A relationship between “Class” and “Functional-
<owl:Class rdf:ID=”Course”>
<owl:DatatypeProperty rdf:ID=”courseNo”>
<!-- NOT NULL -->
<rdfs:domain rdf:resource=”Course”/>
<rdfs:range rdf:resource=”&xsd;short”/>
Then we have “IIR(Class, DatatypeProperty)”, for
example, IIR(Course, courseNo).
Furthermore, to handle a many-to-many relationship,
we transform it into two many-to-one relationships. Con-
sider firstly this scenario: “one course is taken by more
than one students and one student takes more than one
course”. This is a many-to-many relationship. We de-
compose such a relationship into two many-to-one by
creating a new class and then they are treated in the same
way as the second method for the ObjectProperty trans-
formation. We create an intermediate table and use the
ObjectProperty name as the new class name (as well as
the table name which will be the transformation of this
class). In details, the relationship “StudentTakeCourse”
between the class “student” and the class “course” is
many-to-many. We create a new class CourseLearning,
which contains two ObjectProperty relationships as
shown below:
Class (Student)
DatatypeProperty (studentNo domain (Student) range
(xsd: short) Functional)
DatatypeProperty (studentName domain (Student)
range (xsd: string))
DatatypeProperty (major domain (Student) range (xsd:
DatatypeProperty (enrollmentDate domain (Student)
range (xsd: date))
Class (Course)
DatatypeProperty (courseNo domain Course) range
(xsd: short) Functional)
DatatypeProperty (courseName domain (Course)
range (xsd: string))
DatatypeProperty (creditHour domain (Course) range
(xsd: integer))
Class (CourseLearning)
ObjectProperty (takenBy domain (CourseLearning)
range (Student))
ObjectProperty (inv - takenBy domain (Student) range
(CourseLearning) inverseOf (takenBy))
ObjectProperty (takesCourse domain (CourseLearning)
range (Course))
ObjectProperty (inv - takesCourse domain (Course)
range (CourseLearning) inverseOf (takesCourse))
Accordingly the IIR obtained in this process are
IIR(CourseLearning, Student) and IIR(CourseLearning,
Course) (Figure 4).
The paragraphs that follow illustrate details at the “in-
stances” (data values) level of the above example.
The original class CourseLearning is divided into two
parts takenBy and takesCourse (as ObjectProperty), ei-
ther of which only shows a one-way relationship. These
combined however form the relationship between stu-
dents and courses. Table 1 shows some instances.
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Figure 4. The overview for IIR relationship of CourseLearning
Table 1. Table transformation from CourseLearning
CourseLearning Student Course
d1 S1 C1
d2 S2 C1
d3 S1 C2
d4 S4 C1
d5 S3 C2
d6 S4 C2
3.2 Deriving IIR from a Database
In a database, initial IIR (i.e., IIR that is not implied by
others) come from three sources: relationships between
tables, relationships between attributes and relationships
between individual data values. Two different ways can
be used to derive such IIR.
A relationship between a “subclass” and a “super
Figure 5 shows part of a “university” database schema.
An IIR exists between two tables if one is a super class
of the other, for example, IIR(postgraduate, student).
Note that, IIR is a relationship between events as we
said earlier. For tables, we define events as follows: if we
randomly chose a tuple from a database, that the tuple
happens to be in a particular table is an event. Thus the
above IIR(postgraduate, student) means that the exis-
tence of a tuple in table Postgraduate makes certain that a
tuple that corresponds to the former exists in table Stu-
A many-to-one relationship between two tables
Similar to deriving IIR from an “ObjectProperty” with
an ontology, we obtain IIR(table 1, table 2) if they have a
many-to-one relationship for example, IIR(undergraduate,
Constraints of the relational data model and busi-
ness rules on data
A third source of IIR is constraints of a relational da-
tabase and business rules on data, for example, IIR(table
1, PK) and ‘IIR(PK, attribute1). For Figure 5, we have
IIR(courses, courseID) and IIR(courseID, courseName).
The former means that the existence of a tuple of table
Courses makes certain that a corresponding course ID (a
value) exists. The latter means that the existence of a
course ID makes certain that a corresponding course
name exists.
Another type of IIR is IIR(FK1, PK1), for example:
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Figure 5. 6 partial EER diagram of “University” relational database
IIR(courseIDoftableStudents, courseIDoftableCourses),
which means that the existence of a course ID in table
Students makes certain that a corresponding course ID
exists in table Courses.
A database is normally populated with data values.
Inaddition to the above IIR on the “table” level and “at-
tribute” level, there could be IIR identifiable at the “data
value” level.
The information that each individual “data value”
holds in a relational database comes from the semantics
of the attributes to which the data value belongs. This is
due to the capacity of a concept’s “giving meaning to its
instances” [9]. An attribute may be seen as representing a
concept. For example, “student name” is seen as a con-
cept. Relationships between entities in a database can be
seen as “complex concepts” [9] and therefore also give
meaning to data values that are instances of the relation-
ships. That is, data in the relational database already hold
relationships upon which there are constraints imposed.
We now use a simple example to summarise how IIR
may be derived on the three levels. Suppose three tables
shown in Figure 6 in the “University” database.
Table level
According to Figure 6, table administration staff is a
subclass of table staff, which gives the following IIR
between these two tables:
IIR(administration_staff, staff)
As previously mentioned, the meaning of IIR indicates
that first arguments existence results in the certainty that
the other exists, and without the former, the latter is not
certain. Therefore, the meaning of the two part relation-
ship in this particular IIR, may be explained as: “if there
is a member of administration staff then a corresponding
member of staff must exist, otherwise the latter is not
certain”. In this particular case, the IIR is true because
any member of administration staff is a member of staff.
Attribute level
If an attribute “A” is in a table which includes attrib-
utes “A”, “B” and “C”. Then, any combination of “A”,
“B” and “C” that includes “A” would have “A” in its
information content. For example, IIR(AB, A), which
means that if an instance, say (a, b), of “AB” exists,
then there must be an corresponding instance of “A” ex-
isting - in this case, it is a. In general, this type of IIR is
IIR(“a set of attributes”, “a subset of the attributes”).
Using the values in Figure 6, an example is shown
below. The attributes in table administration staff include
sno, position, and deptNo. So the IIR are:
IIR(snopositiondeptNo, sno)
IIR(snopositiondeptNo, snoposition)
IIR(snopositiondeptNo, sno deptNo)
IIR(snopositiondeptNo, position)
IIR(snopositiondeptNo, positiondeptNo)
IIR(snopositiondeptNo, deptNo)
These IIR are derived on the attributes level, which
may be seen as based on the aforementioned “Product”
Rule, i.e., if an event X is the conjunction of a number of
events, then any of the latter is in the information content
of the former.
Data value level
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
sno fnamelname sexAddress tel office
s02923 John Key M 6 Lawrence St, Glasgow2384 E110
s02933 Julie Lee F 8 George St, Glasgow 2234 G203
s04885 Ann White F 18 Taylor St, Glasgow 5112 G133
s04995 SusanBrand F 28 High St, Paisley 3001 G229
s06465 Mary TregearF 7 George St, Paisley 7754 F232
s06883 DavidFord M 64 Well St, Paisley 8772 F231
DB-administration staff
sno position deptNo
s04885 secretary d01
s04995 accountant d03
deptID departmentName
d01 administration
d03 finance
Figure 6. Three tables within the “University” relational database
Unlike the unpopulated ontology in use for this project,
data values are a very important part in relational data-
bases and it also the largest constituent of a relational
database. Before explaining how to derive IIR on the
data value level, let us re-cap the meaning of the terms
we have been using, i.e., “random variable” “event” and
“particular of an event”.
A random variable is an entity used mainly to describe
chance and probability in a mathematical way. An event
is a set of outcomes (a subset of the sample space) to
which a probability is assigned. Typically, when the
sample space is finite, any subset of the sample space is
an event (i.e., all elements of the power set of the sample
space are defined as events) (A WorldViewer.com, 2009).
Moreover, a specific event at a particular time and in a
particular space is called a particular of an event. For
example, consider the following situation. For an electric
circuit, two random variables can be identified: one is
Table 2. IIR between data values
Random variables IIR
iir('s04885', 'secretary')
'sno', 'position'
iir('s04995', 'accountant')
iir('s04885', 'd01')
'sno', 'deptNo'
iir('s04995', 'd03')
“the condition of the lamp”, and the other “the condition
of the switch”. There are two states about the lamp: “lit”
and “unlit”, and two for the switch: “closed” and “open”.
There are 22 events for either. Moreover, “unlit” at 10:30
am and “lit” at 10:30 pm, are two particulars.
Table 2 shows some random variables and associated
for the university database given in Figure 6. In a rela-
-tional database, an attribute, e.g., sno, can be seen as a
random variable, and then each possible data value in
this column is an event. That is, randomly picking up a
tuple in this column, then its value could be any one of
all those that are allowed. An attribute is therefore a
variable. The variable holding a particular value is an
3.3 Deriving IIR from Business Rules
Business rules are domain dependent, established by an
individual organisation and they are ad hoc logical limi-
tations on data. Business rules may be embedded in an
ontology and also could be in a database. In order to de-
rive IIR from these business rules, we treat an Object-
Property in an OWL ontology as if it were a constraint in
a database. Both could be represented as additional rela-
tionships. For example, in a university, there might be a
rule: “Any newly recruited lecturer must hold a PhD”.
Then we have an IIR(newly recruited lecturer, PhD),
which means that if someone is a newly recruited lecturer,
then she/he must hold a PhD corresponding to him/her.
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
4. Testing our Idea
In this section, we show a case study that verifies our
idea. We created an ontology entitled “Academic” and a
relational database “University”. The program runs in
Oracle using PL/SQL. This case study elucidates the dif-
ferent results when reasoning is based on the database in
question only, and on both the database and the ontology
integrated through IIR.
There are 11 tables in the “University” database shown
in Figure 7.
And as we mentioned in previous sections, in the
OWL ontology, a “Class” is transformed into a “table”.
“subClassOf” is treated as a “Class”. A “DatatypeProp-
erty” is changed to an “attribute”. An “ObjectProperty” is
a relationship between classes, which are transformed
into constraints upon these tables. So, we arrive at 10
tables shown in Figure 8 from the “Academic” ontology.
Thus the schema of the “University” is substantially
extended as shown in Figure 9, from which more IIR are
Figure 10 shows the 10 tables in SQL Plus of Oracle.
4.1 Original IIR and Derived IIR
A few business rules are defined for this case study. They
specify correspondences between the “Academic” OWL
ontology and the “University” relational database. For
example, there is a table in the “University” database
called “staff”. There is a class in the “Academic” OWL
ontology named “Person”, and “staff” is a subclass of
“Person”. Other 18 business rules are concerned with
equivalent classes between ontology and the database at
class level. There is one on the ObjectProperty.
The original IIR derived from the ontology and the
‘database and business rules are shown in Table 3.
Applying the IIR inference rules listed earlier to the
IIR identified, more IIR are derived.
4.2 Implementation and Results
As we previously mentioned, firstly, we created tables
for the relational database (named 0db.sql), the ontology
(named 0onto.sql) and IIR (named 0IIR.sql), used for
storing both original IIR from the ontology and the data-
base and 0IIR_DB.sql is used for storing the original IIR
from the relational database alone.
Secondly data values are inserted into the database ta-
bles and all the original IIR are entered into the IIR tables.
Then all single attributes and class names and attribute
names from the ontology and the database are obtained
and inserted into the attributes table with duplicate com-
ponents removed.
Thirdly, three intermediate tables are created. The ta-
bles named fo1 and fo2 store the former and the latter
part of original IIR respectively, and the table t1 stores
intermediate results.
Fourthly, a procedure is invoked. For instance, when a
user asks a question, relevant IIR closures will be looked
at. They embody relevant information for the query.
There is a string match function in this procedure.
In order to find out the difference that the ontology
makes, we compare the two results. One was obtained by
using both the “Academic” OWL ontology and the
“University” database, and the other obtained using the
database alone. They are shown in Table 4.
As Table 4 shows, the first column “Attributes” indi-
cate all attributes that are extracted from “Academic”
OWL ontology and the “University” database. The col-
umn “Closures from both ONTO and DB” shows the IIR
closures that we derive by running our prototype when
the “Academic” OWL ontology is involved, The column
“Closures from DB only” consists of the IIR closures that
we derive by running our prototype when only the “Uni-
versity” database is involved.
We use the same five questions in the testing. As Ta-
ble 4 shows, the attributes that are included in the results
are ticked. When the database alone is used, a query for
“sno” gives 12 results and a query for “matricNo” gives
7 results. When both the ontology and the database are
used, the same query for “sno” gets 14 results and the
same query for “matricNo’ gives 9 results. That is, two
more attributes are found to be included in the respective
IIR closures when the ontology is involved, which means
that more information is made available due to the on-
5. Conclusions
We have described how an ontology may be linked with
database in order to derive hidden information. A proto-
type in Oracle was developed to verify our ideas. We use
the notion of IIR (Information content Inclusion relation)
and inference rules for IIR.
We have found that if we do not invoke a relevant on-
tology, a query may be unanswerable. After invoking an
ontology, more relationships between objects become
available, and therefore more elements can connect to
one another, and as a result, a query may become answer-
able, and as a result, more information can be derived
from data in a database. To achieve this, a key is to be-
able to identify IIR from both a database and an ontology.
We have presented a way of doing so.
More work need to be done in the future, for instance,
to display correspondences between a query and the an-
swers in a more accurate and specific way, i.e., not just
listing the answers. One issue that is not aesthetic is how
to achieve semantic alignment between an ontology and
a database, on which we are currently working.
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
sno fname lname sex Address tel
s02923 John Key M 6 Lawrence St, Glasgow 2384
s02933 Julie Lee F 8 George St, Glasgow 2234
s04885 Ann White F 18 Taylor St, Glasgow 5112
s04995 Susan Brand F 28 High St, Paisley 3001
s06465 Mary Tregear F 7 George St, Paisley 7754
s06883 David Ford M 64 Well St, Paisley 8772
1. DB-staff
sno title school sno school pno
s02923 lecturer computing s06465 business p00203
s02933 professor business s06883 engi-
2. DB-faculty 3. DB-researcher
sno position deptNo matricNopno
s04885 secretary d01 ts030283 p00334
s04995 accountant d03 tm051083p00203
4. DB-administration staff 6. DB-postgraduates
matricNo fname lname sex Address
ts030283 Tony Shaw M 20 George St, Paisley
tm051083 Tina Murphy F 16 George St, Paisley
rn050385 Robert Nielson M 11 George St. Paisley
hf151186 Henry Ford M 7 Well St. Paisley
jw010483 John White M 5 Novar Dr, Glasgow
sb210682 Susan Brand F 2 Manor Rd, Glasgow
cp020381 Chris Paul M 6 Lawrence St, Glasgow
5. DB-student
matricNo creditsSoFar matricNo
rn050385 155 M050385
8. DB-projects 7. DB-undergraduates
courseID courseName creditHour lecturerNo school
c0054 Oracle Development 24 s02923 computing
c0021 International Finance Planning 24 s06465 business
c0154 Advanced Oracle Development 24 s02923 computing
c0155 Networking Principles 16 s06883 computing
c0220 Software Development 24 s06883 computing
9. DB-courses
matricNo courseID results deptID depart-
rn050385 c0054 A d01 administration
hf151186 c0021 C1 d03 finance
cp020381 c0154 B1
cp020381 c0155 C2
sb210682 c0220 B2
10. DB-achievements 11. DB-departments
Figure 7. Tables in the “University” database
matricNo creditsSoFar
rn050385 155
hf151186 65
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
name age sex specialty educationDegree
1. onto-Person 2. onto-Worker
3. onto-Faculty 4. onto-Administration staff
school background supervisor credits
5. onto-Assistants 7. onto-Postgraduates 8. onto-Undergraduates
6. onto-Student
projectNo (PK) projectName courseNo (PK) courseName creditHour
9. onto-Projects 10. onto-Course
Figure 8. Table transformations from the “Academic” ontology
Figure 9. “University” database extended due to an ontology
school title background department Position background
(PK) studentName major address E-mail sex
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Figure 10. The “Academic” ontology represented in SQL Plus
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Table 3. IIR derived from the “Academic” OWL ontology and the “University” relational database
IIR derived from the ‘Academic’ OWL
IIR derived from the ‘University’ Relational Database IIR derived from Business Rules
(corresponding relations)
class, subclass (17) and equivalent class
class, subclass (5) and equivalent class (0) class, subclass (1) and equivalent
class, attributes (20)
1.IIR(Worker, Person) 1.IIR(faculty, staff) 1.IIR(staff, Person)
2.IIR(Faculty, Worker) 2.IIR(administration_staff, staff) 2.IIR(Worker, staff)
3.IIR(Professors, Faculty) 3.IIR(researcher, staff) 3.IIR(Faculty, faculty)
4.IIR(Lecturer, Faculty) 4.IIR(postgraduates, student) 4.IIR(Administration_staff, admini-
5.IIR(Postdoc, Faculty) 5.IIR(undergraduates, student) 5.IIR(Projects, projects)
6.IIR(Administration_staff, Worker) 6.IIR(Student, students)
7.IIR(Dean, Administration_staff) 7.IIR(Postgraduates, postgraduates)
8.IIR(Chair, Administration_staff) 8.IIR(Undergraduates, undergraduates)
9.IIR(Clerical_staff, Administration_staff) 9.IIR(Course, courses)
10.IIR(System_staff, Administration_
10.IIR(courseNo, courseID)
11.IIR(Director, Administration_staff) 11.IIR(projectNo, pno)
12.IIR(Assistants, Worker) 12.IIR(staff, Worker)
13.IIR(Reacher_assistants, Assistants) 13.IIR(faculty, Faculty)
14.IIR(Teaching_assistants, Assistants) 14.IIR(administration_staff, Ad-
15.IIR(Student, Person) 15.IR(projects, Projects)
16.IIR(Postgraduates, Student) 16.IIR(students, Student)
17.IIR(Undergraduates, Student) 17.IIR(postgraduates, Postgraduates)
18.IIR(Teachers, Faculty) 18.IIR(undergraduates, Undergradu-
19.IIR(courses, Course)
20.IIR(courseID, courseNo)
According to the ‘University’ relational database EER
diagram (Figure 8), these 5 IIR could be derived from it.
21.IIR(pno, projectNo)
ObjectProperty (7) and equivalent Ob-
jectProperty (2)
ObjectProperty (5) and equivalent ObjectProperty (0) ObjectProperty (0) and equivalent
ObjectProperty (3)
1.---teache_of IIR(Faculty, Course) 1.---work_in IIR(administration_staff, departments 1.IIR(has, teache_of)
2.---attend_course IIR(Student, Course) 2.---has IIR(faculty, courses) 2.IIR(research_in, work_on)
3.---research_by IIR(Professors, Projects) 3.---employed_on IIR(researcher, projects) 3 IIR(study_in, work_on)
4.---instruct_by IIR(Course, Faculty) 4.---work_on IIR(postgraduates, projects)
5.---research_in IIR(Postgraduates, Pro-
5.---take IIR(undergraduates, courses)
6.---study_in IIR(Postgraduates, Projects)
7.---join_course IIR(Student, Course)
8. IIR(study_in, research_in)
9.IIR(join_course, attend_course)
Constraints----NOT NULL (6) constraints----PK (22) constraints (0)
1.IIR(studentNo, ‘studentNo, student-
1.IIR(sno, ‘sno,fname,lname,sex,address,tel,office’)
2.IIR(courseNo, ‘courseNo, courseName,
2.IIR(sno, ‘sno,title,school’)
3.IIR(projectNo, ‘projectNo, projectName’) 3.IIR(sno, ‘sno,school,pno’)
4.IIR(Student, studentNo) 4.IIR(sno, ‘sno,position,deptNo’)
5.IIR(Projects, projectNo) 5.IIR(matricNo, ‘matricNo,fname,lname,sex,address’)
6.IIR(Course, courseNo) 6.IIR(matricNo, ‘matricNo,pno’)
7.IIR(matricNo, ‘matricNo,creditsSoFar’)
8.IIR(pno, ‘pno,projectName’)
9.IIR(courseID, ‘courseID, courseName, creditHour, lec-
turerNo, school’)
10.IIR(‘matricNo,courseID’, ‘matricNo,courseID,results’)
11.IIR(deptID, ‘deptID,departmentName’)
12.IIR(staff, sno)
13.IIR(faculty, sno)
14.IIR(researcher, sno)
15.IIR(administration_staff, sno)
16.IIR(student, matricNo)
17.IIR(postgraduates, matricNo)
18.IIR(undergraduates, matricNo)
19.IIR(projects, pno)
20.IIR(courses, courseID)
21.IIR(achievements, ‘matricNo,courseID’)
22.IIR(departments, deptID)
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
Table 4. IIR closures compared
Attributes Closures from both ONTO and DB
(the number of results)
Closures from DB only
(the number of results)
Worker(3) faculty(25) student(22) sno(14)matricNo(9)Worker(1)faculty(16)students(1) sno(12) matricNo(7)
6. Acknowledgements
This work is partly sponsored by the a grant for Distrib-
uted Information Systems Research from the Carnegie
Trust for Universities of Scotland, 2007, a grant for re-
search on Semantic Interoperability between Distributed
Digital Museums from the Carnegie Trust for Universi-
ties of Scotland, 2009, and a PhD studentship of the
University of the West of Scotland, UK.
[1] T. R. Gruber, “A Translation Approach to Portable On-
tologies,” Knowledge Acquisition, Vol. 5, No. 2, 1993, pp.
[2] M. West, “Database and Ontology [Online],” 2008. Wiki
HomePage. http://ontolog.cim3.net/cgi-bin/wiki.pl?Data
[3] T. Berners-Lee, J. Hendler and O. Lassila, “The Semantic
Web,” Scientific American, Vol. 284, No. 5, 2001, pp. 34-
[4] Z. M. Xu, S. C. Zhang and Y. S. Dong, “Mapping be-
tween Relational Database Schema and OWL Ontology
for Deep Annotation,” Proceedings of the 2006 IEEE/
WIC/ACM International Conference on Web Intelligence,
IEEE Computer Society, 2006, pp. 548-552. http://por-
Using an Ontology to Help Reason about the Information Content of Data
Copyright © 2010 SciRes. JSEA
[5] C. B. Necib and J. C. Freytag, “Query Processing Using
Ontologies,” Proceedings of 17th International Confer-
ence on Advanced Information Systems Engineering,
Springer, Porto, Portugal, 13-17 June 2005.
[6] K. Munir, M. Odeh and R. McClatchey, “Ontology As-
sisted Query Reformulation Using the Semantic and As-
sertion Capabilities of OWL-DL Ontologies,” Proceed-
ings of the 2008 International Symposium on Database
Engineering & Applications, ACM, Coimbra, Portugal,
2008, pp. 81-90.
[7] J. Feng, “The ‘Information Content’ Problem of a Con-
ceptual Data Schema,” Systemist, Vol. 20, No. 4, 1998,
pp. 221-233.
[8] K. Xu, J. Feng and M. Crowe, “Defining the Notion of
‘Information Content’ and Reasoning about it in a Data-
base,” Knowledge and Information Systems, Vol. 18, No.
1, 1 January 2009, pp. 29-59
[9] F. I. Dretske, “Knowledge and the Flow of Information,”
MIT Press, Cambridge, 1981.
[10] K. Loney, “Oracle Database 10g: The Complete Refer-
ence,” McGraw-Hill Companies, Inc., NY, 2004.
[11] B. C. Grau and B. Motik, “OWL 1.1 Web Ontology Lan-
guage: Model-Theoretic Semantics. W3C Working Draft
[Online],” 8 January 2008. W3C. http://www.w3.org/TR/
[12] Z. M. Xu and Y. J. Huang, “Conversion from OWL On-
tology to Relational Database Schema,” College of
Computer and Information Engineering, Hohai University,
Nanjing, 2006.