=Paper=
{{Paper
|id=Vol-1193/paper_25
|storemode=property
|title=Transition Constraints for Temporal Attributes
|pdfUrl=https://ceur-ws.org/Vol-1193/paper_25.pdf
|volume=Vol-1193
|dblpUrl=https://dblp.org/rec/conf/dlog/OngomaKM14
}}
==Transition Constraints for Temporal Attributes==
https://ceur-ws.org/Vol-1193/paper_25.pdf
Transition Constraints for Temporal Attributes
E.A. Nasubo Ongoma1 , C. Maria Keet2 , and Thomas Meyer1
1
School of Mathematics, Statistics, and Computer Science, University of KwaZulu-Natal and
UKZN/CSIR-Meraka Centre for Artificial Intelligence Research, South Africa,
212562258@stu.ukzn.ac.za, tmeyer@csir.co.za
2
Department of Computer Science, University of Cape Town, South Africa,
mkeet@cs.uct.ac.za
Abstract. Representing temporal data in conceptual data models and ontologies
is required by various application domains. For it to be useful for modellers to
represent the information precisely and reason over it, it is essential to have a lan-
guage that is expressive enough to capture the required operational semantics of
the time-varying information. Temporal modelling languages have little support
for temporal attributes, if at all, yet attributes are a standard element in the widely
used conceptual modelling languages such as EER and UML. This hiatus pre-
vents one to utilise a complete temporal conceptual data model and keep track of
evolving values of data and its interaction with temporal classes. A rich axioma-
tisation of fully temporised attributes is possible with a minor extension to the
already very expressive description logic language DLRU S . We formalise the
notion of transition of attributes, and their interaction with transition of classes.
The transition specified for attributes are extension, evolution, and arbitrary quan-
titative extension.
1 Introduction
Representing temporal information has been researched for over 20 years in diverse
fields, including temporal conceptual models [17, 1, 16], logic-based representation [1,
4], informal business rules on time [12], spatio-temporal models [18] and Ontology
Based Data Access [2, 6]. Logic-based representation of temporal data at the concep-
tual level provides a link between temporal conceptual data models and temporal de-
scription logic. One of the notable achievements is the temporally extended ER lan-
guage called ERV T that has a logical reconstruction in the description logic language
DLRU S , which covers temporal behaviour of classes [1, 3] and relationships [14] and
more recently also temporal attributes [15].
According to [11], temporal modelling addresses the question how does the world
change? One way is as the life cycle, the evolving state of membership of an object
from its start (creation) to end (deletion), covered in [3, 14, 15] for objects, relations and
attributes respectively. The second way is transition, concerning the change of object
properties over time as the object migrates from one class to the next. Research on
transition constraints has covered only object migration [3] and relation migration [14].
However, to efficiently represent and reason over data, we need to also look at attribute
transition, which is the migration of attributes along an object’s life cycle. Temporal
attributes have been used and studied less in comparison to classes and relations. Yet in
databases, we have a direct contact with attributes, and if those (temporal) attributes are
not represented properly in our conceptual models, one will face inconsistent databases
with respect to the constraints that ought to hold.
This paper looks at temporal attributes as they evolve over time along an object’s
lifecycle. The formalisation is done using a minor extension of the temporal description
logic DLRU S with temporal attributes. This work relies on previous results on transi-
tions of objects and relations for ERV T . Using the constraints described in the previous
work, we represent attribute migration as evolution constraints A DEV and A DEX, per-
sistence constraints A PEX and A PEV and quantitative constraints AQEV and AQEX. We
also give a more rigorous definition to arbitrary quantitative transition of attributes in
time (instead of merely stating the next point) in time.
The paper is organised as follows. We first introduce a few motivating scenarios for
temporal attributes in Section 2. The basics of transition and the modified description
logic DLRU S is described in Section 3. The new semantics for quantitative evolution as
well as the transition constraints for attributes, with its interaction with evolving classes
is described in Section 4. We discuss the results in Section 5 and conclude in Section 6.
2 Application Areas
Attribute migration is done in practice but it is not explicitly stated and formally rep-
resented. Our aim is to introduce the notion of evolving temporal attributes to stake-
holders of the various application domains. Most businesses have as aim to reduce cost-
to-company and by using transition constraints companies would reduce cost as well
as eliminate human error in their databases. Transition constraints also permit com-
panies to represent history and learn from it, for example policy makers in insurance
companies make their decisions based from the client’s history, reservation systems use
temporal databases to manage booking and medical information systems to manage
patient’s records. Evolving temporal attributes can therefore be used in many applica-
tion areas mapped as business rules to monitor, verify and ensure that the processes
run as planned and also check consistency, by reflecting violations and generate new
knowledge. System administrators would then have to write down the requirements for
data-centric transitions and the objects and attributes in database would evolve after the
set amount of time. Consider the following application areas:
– Security features put in place to prevent a breach in access and ensure safety and
dependability, for example if an employee is promoted to a branch manager but he
is given the access rights of a CEO due to underspecified change constraints on
the data, then there would be a violation. The attributes can be given according to
the employee ranks, (say from 1-6), let’s say that an employee is given the lowest
access, different managers with different access rights and the highest is the CEO.
– Administration would benefit from human resource business rules that permit and
manage evolution of classes. For example having the promotion of an employee,
if we had a business rule that all managers must have a masters degree and an
employee is promoted without one, it will cause a violation in the database.
– Medical information systems to monitor the effect drugs on patients and the sta-
tus of the patients. For instance, there are sensor readings for a particular patient
and when it reaches a particular state, medical intervention is required, for exam-
ple when a diabetes out-patient’s blood sugar reaches a certain level, he needs to
be hospitalised. It is not trivial to declare evolving constraints in the medical and
healthcare field, but it is possible to monitor patients’ changes (e.g., in the ICU),
recorded with a set of attributes, and show evolution of the patient’s states against
the current state, raise alarms on extreme changes, and reclassify the patient ac-
cording to the symptoms.
– Situational awareness, to monitor the changes in an environment and add meaning
to this change. For example, recording the change in the levels (concentration) of a
toxin in the environment, or even the change in climate zone designation of a region
(recorded with attributes in a GIS).
The above examples show how transition constraints of attributes are used and why we
need to properly represent them in order to capture the operational semantics governing
the movement and change of data in a temporal conceptual model or ontology.
3 Preliminaries
We briefly introduce what has been done before on the life cycle, status classes and tran-
sitions. The life cycle can be viewed in three ways: by status classes [3], status relations
[14] and status attributes [15] which records the membership of an object, relation, or
attribute as it evolves along the statuses. Status models the normal behaviour of the life
cycle in the real world, items are first scheduled, then become active, can be suspended
and reactivated again and finally may become disabled, when they reached their expiry.
These are the four statuses used when modelling data, which are represented in an EER
diagram as shown in Fig. 1; the rectangular boxes are status classes (entity types) and
the ovals are status attributes, which are ‘housed’ in classes. An active class (C) can
have any of the four status attributes, while scheduled, suspended and disabled classes
can only have their corresponding attributes, i.e. scheduled, suspended or disabled only.
Transition records migration of elements from source to target element, be they
objects or relations as they move along their lifecycle. They control permissible changes
from source class to target, to ensure that data does not enter an impossible state because
of a previous state. These objects (relations) may lose membership from the source class
(relation) or acquire other memberships in the target class (relation), which may give
rise to a new classification scheme. Classes have both temporal and snapshot attributes,
but only temporal attributes can participate in the migration, whereas snapshot attributes
always remain the same.
Object transitions give constraints that capture the movement of an object from
source class to target class. These transitions can be either an evolution or an extension,
the difference being that an object in an evolution ceases to be a member of the source
class while in an extension, the object is still a member of the source class. Dynamic
evolution DEV and dynamic extension DEX were introduced in [3], while quantitative
extension TEX, quantitative evolution TEV, persistent extension PEX, persistent evolu-
tion PEV were introduced in [5]. The parallel transition for relations are relation dy-
Top C
d
Exists-C spev Disabled-C
d
dev
Scheduled-C sdev C dev Suspended-C
Top A
d
Exists-A Disabled-A
sapev
d
adev
Scheduled-A A Suspended-A
sadev adev
Fig. 1. EER diagram with the class hierarchy of status classes (rectangles) integrated with the
attribute hierarchy of status attributes (ovals). The dashed arrows illustrate class migration and
attribute migration.
namic evolution R DEV, dynamic extension for relations R DEX, quantitative extension
for relations R QEX, quantitative evolution for relations R QEV, strong dynamic exten-
sion for relations SR DEX and strong dynamic extension for relations SR QEX transition
relations were introduced in [14]. However, there is no support for attribute migration
or on the effect of object migration on attributes, yet every object transition affects
attributes.
3.1 DLRU S with Attributes
The Description Logic DLRU S [1] is an expressive fragment of FOL that combines
the propositional temporal logic with Since and Until operators with the (non-temporal)
description logic DLR [9] so that relationships and classes can be temporalised. We add
a minor extension to DLRU S to include a precise syntax and semantics for attributes,
mainly because we are dealing with conceptual data models that require attributes. A
temporary attribute was defined earlier in [3], as a binary relation for each attribute,
A ∈ A, for hA, Ci ∈ T , with its DLRU S axiom as C v ¬∃[From](2∗ A). This entailed
that temporal constructors can be used in front of attributes, which, however, was not
included in the DLRU S syntax and semantics.
The syntax and semantics of the extended DLRU S are included in Fig. 2. Details
of DLRU S can be found in [1, 3]; as usual, we have classes C (starting from atomic
ones CN ), n-ary relations R (DL roles, with n ≥ 2, RN ), binary attributes A between
a class and a datatype, DL role components (U , of which F denotes a role component
in an attribute, F ⊆ U, and F = {From, To}). The selection expression Ui /n : C
denotes an n-ary relation whose i-th argument (i ≤ n) is of type C and [Uj ]R denotes
the j-th argument (j ≤ n)—DL role component, which can be seen intuitively as a
projection over the role—in role R (subscripts i and j are omitted if it is clear from
the context). For F , which concerns the DL role components in an attribute, we thus
have F : C, with F denoting the role component From that relates to class C, and [F ]A
denoting the role component F of A, where if To is used, it is the DL role component
that associates with the datatype of the attribute, and if From is used, it is the DL role
component that associates with the class of the attribute. Thus, for each A ∈ A and
denoting with Literal the top for data types (i.e., for the domain of values ∆ID ; see
below), the DLRU S axiom A v From : > u To : Literal holds. Finally, Until and
Since together with ⊥ and > suffice to define the temporal operators: 3+ (some time
in the future) as 3+ C ≡ > U C, ⊕ (at the next moment) as ⊕ C ≡ ⊥ U C, and likewise
for their past counterparts; 2+ (always in the future) and 2− (always in the past) are
the duals of 3+ and 3− .
The model theoretic semantics of DLRU S assumes a flow of time T = hTp , <
i, where Tp is a set of countably infinite time points also referred to as chronons and < is
isomorphic to the usual ordering on the integers. The language of DLRU S is interpreted
in temporal models over T , which are triples in the form I = hT , ∆I , ·I(t) i, where ∆I
is the union of two non empty disjoint sets, the domain of objects, ∆IO , and domain of
values, ∆ID , and ·I(t) the interpretation function such that, for every t ∈ T (t ∈ T will
be used as a shortcut for t ∈ Tp ), every class C, and every n-ary relation R, we have
C I(t) ⊆ ∆IO and RI(t) ⊆ (∆IO )n ; also, (u, v) = {w ∈ T | u < w < v}. Note that in
[1, 3], ∆ID was already used in the DLRU S -based logic reconstruction of ERV T , and
∆I = ∆IO ∪∆ID , but this was not explicitly stated in the DLRU S syntax and semantics;
it is now in Fig. 2.
A knowledge base is a finite set Σ of DLRU S axioms of the form C1 v C2 and
R1 v R2 , and with R1 and R2 being relations of the same arity. An interpretation I
satisfies C1 v C2 (R1 v R2 ) if and only if the interpretation of C1 (R1 ) is included
I(t) I(t) I(t) I(t)
in the interpretation of C2 (R2 ) at all time, i.e. C1 ⊆ C2 (R1 ⊆ R2 ), for all
t∈T.
4 Formalising Transition of Attributes
Status attributes are introduced in [15], which reuse the idea of status classes from [3],
and its hierarchy is shown in Fig. 1. It has a rigorous formalisation that specifies the
operational semantics of temporal attributes, to control the movement and permissible
transitions of attributes, though we commit to the intention of suspension, such that
suspended classes cannot participate in a transition. Previous formalisation of evolution
constraints DEX and DEV in [3] asserted that both suspended classes and active classes
can participate in a transition. This violates its ISA2 proposition (objects suspended in a
subclass must also be suspended or active in the superclass), which, if permitted, would
allow other operations on suspended entities and to participate in transitions, e.g., one
should not permit an extension of an object from a Suspended-Employee to an Active-
Manager if Manager v Employee. The class must first be active before any transition.
In addition, we add the notion of arbitrary quantitative evolution going beyond one
chronon (the next point (t + 1)). The essence here is to have a way to represent fixed
C→ > | ⊥ | CN | ¬C | C1 u C2 | ∃≶k [Uj ]R | ∃[F]A |
3+ C | 3− C | 2+ C | 2− C | ⊕ C | C | C1 U C2 | C1 S C2
R→ >n | RN | ¬R | R1 u R2 | Ui /n : C |
3+ R | 3− R | 2+ R | 2− R | ⊕ R | R | R1 U R2 | R1 S R2
A→ >A | AN | ¬A | F : C |
3+ A | 3− A | 2+ A | 2− A | ⊕ A | A | A1 U A2 | A1 S A2
>I(t) = ∆I O
⊥I(t) = ∅
I(t) I(t)
CN ⊆ >
(¬C)I(t) = >I(t) \ C I(t)
I(t) I(t) I(t)
(C1 u C2 ) = C1 ∩ C2
(∃≶k [Uj ]R)I(t) = { o ∈ >I(t) | ]{ho1 , . . . , on i ∈ RI(t) | oj = o} ≶ k}
(∃ [F]AI(t) = { o ∈ >I(t) | ]{ho, di ∈ AI(t) ≥ 1}}
I(v) I(w)
(C1 U C2 )I(t) = { o ∈ >I(t) | ∃v > t.(o ∈ C2 ∧ ∀w ∈ (t, v).o ∈ C1 )}
I(v) I(w)
(C1 S C2 )I(t) = { o ∈ >I(t) | ∃v < t.(o ∈ C2 ∧ ∀w ∈ (v, t).o ∈ C1 )}
(>n )I(t) = (∆IO)
n
RN I(t) ⊆ (>n )I(t)
(¬R)I(t) = (>n )I(t) \ RI(t)
I(t) I(t)
(R1 u R2 )I(t) = R1 ∩ R2
(Ui /n : C)I(t) = { ho1 , . . . , on i ∈ (>n )I(t) | oi ∈ C I(t) }
I(v)
(R1 U R2 )I(t) = { ho1 , . . . , on i ∈ (>n )I(t) | ∃v > t.(ho1 , . . . , on i ∈ R2 ∧
I(w)
∀w ∈ (t, v). ho1 , . . . , on i ∈ R1 )}
I(t) I(v)
(R1 S R2 ) = { ho1 , . . . , on i ∈ (>n )I(t) | ∃v < t.(ho1 , . . . , on i ∈ R2 ∧
I(w)
∀w ∈ (v, t). ho1 , . . . , on i ∈ R1 )}
(3+ R)I(t) = {ho1 , . . . , on i ∈ (>n )I(t) | ∃v > t. ho1 , . . . , on i ∈ RI(v) }
(⊕ R)I(t) = {ho1 , . . . , on i ∈ (>n )I(t) | ho1 , . . . , on i ∈ RI(t+1) }
(3− R)I(t) = {ho1 , . . . , on i ∈ (>n )I(t) | ∃v < t. ho1 , . . . , on i ∈ RI(v) }
( R)I(t) = {ho1 , . . . , on i ∈ (>n )I(t) | ho1 , . . . , on i ∈ RI(t−1)
(>A )I(t) = ∆IO × ∆D
I
ANI(t) ⊆ (>A )I(t)
(F : C)I(t) = { ho, di ∈ (>A )I(t) | o ∈ CI(t) }
I(v) I(w)
(A1 U A2 )I(t) = { ho, di ∈ (>A )I(t) | ∃v > t.(ho, di ∈ A2 ∧∀w ∈ (t, v).ho, di ∈ A1 )}
I(v) I(w)
(A1 S A2 )I(t) = { ho, di ∈ (>A )I(t) | ∃v < t.(ho, di ∈ A2 ∧∀w ∈ (v, t).ho, di ∈ A1 )}
I(t)
(3+ A) = {ho, di ∈ (>A )I(t) | ∃v > t.ho, di ∈ AI(v) }
(⊕ A)I(t) = {ho, di ∈ (>A )I(t) | ho, di ∈ AI(t+1) }
I(t)
(3− A) = {ho, di ∈ (>A )I(t) | ∃v < t.ho, di ∈ AI(v) }
( A)I(t) = {ho, di ∈ (>A )I(t) | ho, di ∈ AI(t−1) }
Fig. 2. Syntax and semantics of DLRU S , modified to include attributes (in bold face); o denote
objects, d domain values, v, w, t ∈ T , F is a role component in an attribute.
time changes e.g. transition after 3 years. We provide a generalisation of the next time as
(t + x). We use the subsumption notion that, if we have a class A v ⊕B and B v ⊕C,
we have A v ⊕ ⊕ C, which can also be written in shorthand notation (i.e. syntactic
sugar that does not affect the computational complexity) as A v ⊕2 C. This enables us
to set the number of chronons before a class evolution occurs.
An integrated picture of status classes and status attributes is shown in Fig. 1, the
dashed lines showing transition of objects and attributes.
4.1 Attribute Migration
Attribute migration occurs when an attribute in the same object object migrates to an-
other attribute, for example the attributes has degree and has postgrad, is a transition
from one attribute to another. Attribute migration is more complex than object migration
because it is bidirectional, i.e., it can cause migration of objects, triggering reclassifi-
cation of objects as well as participate in an object migration. For example, an HIV
patient becoming an AIDS patient after his CD4 count (attribute) falls below 180 or a
bank account (class) being frozen, due to expiry of a work permit.
Representation of attribute hierarchies is uncommon, mainly because it is not prop-
erly defined and formalised, but is relevant for modelling temporal data. New results
on attribute hierarchies with subsumption in [15] allow us to capture the interaction
between the permissible statuses of classes and status attributes in temporal transitions.
A proper representation of transition in the conceptual model will enable a modeller to
know how to design a temporal database.
We introduce attribute migration with their axioms in the extended DLRU S , build-
ing up work done on relation migration [14] and object migration [3, 5]. In attribute
migration we look at the values which may change with time. ER diagrams do not
have the capability of representing values on the diagrams but this can be modified by
adding the value alongside the attribute. ORM [13] has “attribute free” diagrams, but
each value type has a mapped to attribute to a datatype, while UML [8] displays the
attribute, datatype, and optionally permitted values inside the class in the diagram.
We discuss attribute transition with their examples and give their description logic
semantics below, ordered along dynamic constraints, quantitative constraints, and per-
sistence constraints. Our assumption is that all the attributes exist in the class and are
active until they are disabled.
1. Dynamic constraints - these constraints model how attribute migration occurs in
an object, from the source attribute to the target attribute. We have three types of
dynamic transitions:
A DEX: Attribute dynamic extension, which occurs when the attributes are still
part of the source attribute as it migrates to the target attribute, for example the
number of degrees you have can only increase over time, from an undergradu-
ate degree to masters to phd. This can be modelled as the attributes has degree
for BSc, and has postgrad for MSc and PhD
(A DEX) Dynamic extension of an attribute
A DEXA1 ,A2 v A1 u ¬A2 u 3+ (A1 u A2 )
A DEV: Attribute dynamic evolution occurs when an attribute migrates from the
source attribute to a target attribute, but ceases to be a member of the source at-
tribute. The attribute changes from the previous one, for example when chang-
ing the name of a product, an attribute new name is added, or for a particular
copyright (a has copyright attribute) on creative work to evolve to public do-
main (pub domain, e.g., as boolean).
(A DEV) Dynamic evolution of an attribute
A DEVA1 ,A2 v A1 u ¬A2 u 3+ (¬A1 u A2 )
S ADEV: Strong attribute dynamic evolution is a subclass of dynamic attribute evo-
lution, which occurs when the source attribute can never go back to the source
attribute, for example a scheduled attribute can never become scheduled after
it has become active.
(S ADEV) Strong Dynamic evolution of an attribute
S ADEVA1 ,A2 v A1 u ¬A2 u 3+ A2 u 2+ ¬A1
2. Quantitative constraints - these constraints specify the exact amount of time an
attribute transition occurs from the source attribute to the target attribute. We use
the generalisation of the next time, (t + x), for quantitative constraints.
AQEX: Attribute quantitative extension occurs when an attribute is set to migrate
after a specified amount of time and the attribute still remains a member of the
source attribute, for example, the modeller may specify that after the proba-
tionary period, say 6 months, a manager starts earning a full salary, whereas
before he earned, say, 90% of his stipulated salary.
(AQEX) Quantitative extension of an attribute
AQEXA1 ,A2 v A1 u ¬A2 u ⊕n (A1 u A2 )
AQEV occurs when attributes migrate after a given time period but are no longer
members of the source attribute, for example, an employee receiving a different
bonus every 2 years.
(AQEV) Quantitative evolution of an attribute
AQEVA1 ,A2 v A1 u ¬A2 u ⊕n (¬A1 u A2 )
S AQEV occurs when attributes migrate after a set period of time and do not go
back to their previous state.
(S AQEV) Strong Quantitative evolution of an attribute
S AQEVA1 ,A2 v A1 u ¬A2 u ⊕n (¬A1 u A2 )
3. Persistence constraints - these constraints specify the persistent, non changing, state
of an attribute after it has migrated from the source attribute to the target attribute.
Once an attribute migrates, its value never changes.
A PEX: Persistent attribute extension occurs when the attribute migrates, but will
never change it value in the future, and is still a member in the source attribute.
(A PEX) Persistent extension of an attribute
A PEXA1 ,A2 v A1 u ¬A2 u 2+ (A1 u A2 )
A PEV: Persistent attribute evolution occurs when the attribute migrates, and ceases
to be a member of the source attribute and its value remains constant from that
point on.
(A PEV) Persistent evolution of an attribute
A PEVA1 ,A2 v A1 u ¬A2 u 2+ (¬A1 u A2 )
S APEV occurs when the attribute migrates but it never changes its value after the
migration. For example when an attribute is disabled, it will never be active, it
will always persist in the disabled state.
(S APEV) Strong Persistent evolution of an attribute
S APEVA1 ,A2 v A1 u ¬A2 u 2+ (¬A1 u A2 )
4.2 Interaction between Transition of Classes and Attributes
Object migration affects only temporal classes and as a result temporal attributes are
affected. The interaction between migrating temporal classes and temporal attributes
brings forth two cases. CASE A in which object migration induces an attribute migra-
tion, and vice versa, CASE B. Although some aspects cannot be formalised in DLRU S ,
it is useful to at least consider the scenarios.
CASE A: The object migration causes an attribute migration such that the attributes
move from the source to the target class. An example for CASE A is illustrated in Fig. 3,
where dashed lines show examples of transitions on the integrated status classes and
status attributes. Whenever a class migrates, the attribute undergoes the same kind of
migration, for instance, when a class ceases to exist, it is disabled, undergoes a dynamic
evolution DEV, which triggers a parallel transition for its attributes, A DEV. For example
when a student ( member of a class Student) graduates and becomes an alumni, his ad-
dress (represented with an attribute) changes and the scheduled attribute of occupation
becomes active. CASE A is not just a byproduct of inheritance because the values of the
attributes change as object migration takes place.
CASE A: CASE B:
Object migration causes attribute migration Attribute migration causes object migration
CD4
eID Employee access pID HIVpos
count
DEX ADEX
ADEX DEX
CD4
eID Manager access pID AIDSpat
count
Fig. 3. EER diagram extended with migration constraints showing the interaction between objects
and attributes. The dashed arrows illustrate class migration and attribute migration.
CASE B: This type of transition occurs when the value of an attribute changes
through an attribute migration and that forces the evolution or extension of an object
from the source to the target class. This case can be very useful in medical information
systems, when monitoring the value of attributes of a patient: if it falls below a certain
threshold, the patient is moved to a different class, which is illustrated in Fig. 3 for HIV
positive to transition to AIDS patient upon passing the threshold of CD4 count of 180. It
can also be used in administration, to group individuals according to a set criteria, for
example when an employee’s annual income increases to above a certain value, he is
moved to a higher class of tax remittance.
4.3 Logical Implications
Logical implications are important to derive new constraints from a set of defined ax-
ioms. Given the set of axioms for attribute migration and the set of axioms for status
attributes [15], several logical implications can be derived. We only need a few axioms
for status attributes for that, being the straight-forward disjointness and completeness
constraints for the four status classes and attributes as indicated in Fig. 1, and using
names C denoting the active C, etc., we have:
(ADISAB 5) Disabled persists, with a semantics as a ∈ Disabled-AI(t) → ∀t0 > t.a ∈ /
0
AI(t ) and in DL-notation Disabled-A v 2+ ¬A;
(ADISAB 4) Disabled will never be Active, Disabled-A v ¬(From : Scheduled-C t
From : Suspended-C);
(C SUSP 3) Freezing attributes of suspended classes, o ∈ Suspended-C I(t) → ho, di ∈
Suspended-AI(t) , and in DL-notation: Suspended-C v ∀[From]Suspended-A;
(A SCH 1) Persists until active, with a semantics a ∈ Scheduled-AI(t) → ∃t0 >t.a ∈
0
AI(t ) and in DL notation Scheduled-A v 3+ A.
Then we can prove the following.
1. Only temporal attributes can participate in the migration
We know that only temporal classes are involved in transition, proven in [3], and
they have both snapshot and temporal attributes. Temporal attributes hold at single
time points only, we prove by contradiction that the attributes cannot be snapshot.
Suppose the attribute is snapshot, A DEXA1 ,A2 v ⊕ A DEXA1 ,A2 u A DEXA1 ,A2 ,
which contradicts because temporal attributes hold only at single time points. There-
fore, we can only have temporal attributes.
2. Only active attributes participate in attribute migration
A temporal active class can have any of the four status attributes, we prove by
contradiction that only active attributes can participate in the transition. To prove,
we use the status attribute axioms from [15], with A defined as ∆IO × ∆ID . Suppose
that
– a ∈ Disabled-AI(t) by (ADISAB 5), but disabled is an irreversible state by
(ADISAB 4). Thus, it can never be modified again, this contradicts, because in a
transition the value changes.
– a ∈ Suspended-AI(t) , by (C SUSP 3) the attributes are frozen, meaning, no
modification can be done on it. This contradicts because the transition warrants
the change of the value
– a ∈ Scheduled-AI(t) By (A SCH 1), and remains unchanged until it is active,
which contradicts.
– a ∈ AI(t) , there is no constraint preventing the change if its value.
5 Discussion
This research provides an avenue for discussion of several issues concerning temporal
data in many application areas and how temporal attributes can contribute to temporal
reasoning. Temporal reasoning can use temporal constraints to check consistency (no
conflicting temporal information) and satisfiability of some constraint, by giving an
accurate representation of reality thus eliminate and filter Transition constraints are
an upcoming research area in OBDA for dynamic objects time points, so that we can
query on the past on historical databases and from that look at how to predict the future.
disallowed (impossible) scenarios. Transition constraints are an upcoming research area
in OBDA for dynamic objects [7] time points, so that we can query on the past on
historical databases and from that look at how to predict the future.
Representing transition constraints for attributes over evolving attributes has not
been researched and we present description logic axioms to describe transition be-
haviour using a minor extended version of the description logic DLRU S for attributes,
as well as some logical implications resulting from the attribute transition. Although
it was possible to formalise several transitions, not all attribute transitions can be rep-
resented, notably attribute transitions that are dependent on one another, and it was
difficult to find relevant use cases for some of them, but not others.
DLRU S does not have a value comparison operator for attributes, needed in order
to formalise CASE B. Due to this limitation, we cannot give the full formalisation, but
it is important to note CASE B as a proposal for future investigation into a suitable
description logic language, and to ensure that we cover all the areas that pertain to
attributes. It may be possible to elaborate on the datatypes as used in OWL 2 [10] or
DL-Lite [4] for more comprehensive logic-based reconstructions of conceptual data
models. We can use transition constraints of attributes to control the behaviour and
movement of objects as they evolve, thus enabling us to verify and validate our data,
using the rules before any evolution occurs, which is a suggested usage also by [19].
DLRU S is undecidable, which might be seen as a drawback. However, it is useful
first to have a language with high expressivity to model what may be needed, rather
than complexity and a priori restricting oneself, which enables a more comprehensive
understanding of temporal attributes and their role in temporal conceptual data mod-
els. In the future, we will look at decidable languages together with a prioritisation of
language constructs needed for the more relevant types of temporal attributes and their
transitions. Based upon that one can choose which of the constraints can be represented
and modelled and used to determine the optimal trade-off between a subset of temporal
constructs and the complexity costs. This research will also permit us to find out the
best way to link up conceptual modelling desires vis-a-vis OBDA [6] to find out what
can be implemented in temporal OBDA.
Research is ongoing in security, which considers fine grained access to data, and
having temporal databases will be needed to ensure correctness of implementation be-
fore there is a build up of possibly dirty data that does not meet the business rules.
Temporal attributes can be used to model time changing data in real world applications
by enabling the creation of trigger rules to manage the integrity of databases. Hopefully,
this paper contributes also to bring awareness on this subject because we need to find
an efficient way to represent temporal attributes in databases that will enable users to
monitor their evolution.
6 Conclusions
Transition constraints are important to record database evolution, which is useful in
several application domains. In this paper we gave a formalisation of evolving temporal
attributes and its interaction with evolving objects using DLRU S .
These results have the potential to assist database designers on how to model tem-
poral data. We are currently looking into designing a graphical modelling tool to test
these results, and future works pertain to the investigation of the best trade-off between
modelling temporal attributes and the complexity of the language.
References
1. Artale, A., Franconi, E., Wolter, F., Zakharyaschev, M.: A temporal description logic for rea-
soning about conceptual schemas and queries. In: Flesca, S., Greco, S., Leone, N., Ianni, G.
(eds.) Proceedings of the 8th Joint European Conference on Logics in Artificial Intelligence
(JELIA’02). LNAI, vol. 2424, pp. 98–110. Springer Verlag (2002)
2. Artale, A., Kontchakov, R., Wolter, F., Zakharyaschev, M.: Temporal description logic for
ontology-based data access. In: Proceedings of the Twenty-Third International Joint Confer-
ence on Artificial Intelligence (IJCAI’13). pp. 711–717 (2013)
3. Artale, A., Parent, C., Spaccapietra, S.: Evolving objects in temporal information systems.
Ann. Math. Artif. Intell. 50(1-2), 5–38 (2007)
4. Artale, A., Ryzhikov, V., Kontchakov, R.: Dl-lite with attributes and datatypes. In: Proceed-
ings of the European Conference on Artificial Intelligence (ECAI’12). pp. 61–66 (2012)
5. Artale, A., Kontchakov, R., Ryzhikov, V., Zakharyaschev, M.: Complexity of reasoning over
temporal data models. In: Parsons, J., Saeki, M., Shoval, P., Woo, C.C., Wand, Y. (eds.) ER.
Lecture Notes in Computer Science, vol. 6412, pp. 174–187. Springer (2010)
6. Baader, F., Borgwardt, S., Lippmann, M.: Temporalizing ontology-based data access. In:
Bonacina, M. (ed.) Automated Deduction – CADE-24, Lecture Notes in Computer Science,
vol. 7898, pp. 330–344. Springer Berlin Heidelberg (2013)
7. Baader, F.: Ontology-based monitoring of dynamic systems (2014)
8. Booch, G., Rumbaugh, J.E., Jacobson, I.: The unified modeling language user guide - the
ultimate tutorial to the UML from the original designers. Addison-Wesley object technology
series, Addison-Wesley-Longman (1999)
9. Calvanese, D., De Giacomo, G.: The DL Handbook: Theory, Implementation and Applica-
tions, chap. Expressive description logics, pp. 178–218. Cambridge University Press (2003)
10. Cuenca Grau, B., Horrocks, I., Motik, B., Parsia, B., Patel-Schneider, P., Sattler, U.: OWL
2: The next step for OWL. Journal of Web Semantics: Science, Services and Agents on the
World Wide Web 6(4), 309–322 (2008)
11. Hall, G., Gupta, R.: Modeling transition. In: ICDE. pp. 540–549. IEEE Computer Society
(1991)
12. Halpin, T.: Temporal modeling and ORM. In: Meersman, R., Tari, Z., Herrero., P. (eds.)
OTM 2008 Workshops. LNCS, vol. 5333, pp. 688–698. Springer (2008)
13. Halpin, T.A.: Object-role modeling: Principles and benefits. IJISMD 1(1), 33–57 (2010)
14. Keet, C.M., Artale, A.: A basic characterization of relation migration. In: Li, Q., Spaccapi-
etra, S., Yu, E., Olivé, A. (eds.) OTM Workshops. LNCS, vol. 5231, pp. 484–493. Springer
(2010)
15. Keet, C.M., Ongoma, E.A.N.: Temporal attributes: their status and subsumption (2014), (in
preparation)
16. Lutz, C., Wolter, F., Zakharyaschev, M.: Temporal description logics: A survey. In: Demri,
S., Jensen, C.S. (eds.) TIME. pp. 3–14. IEEE Computer Society (2008)
17. Object Management Group: Superstructure specification. Standard 2.4.1, Object Manage-
ment Group (2012), http://www.omg.org/spec/UML/2.4.1/
18. Parent, C., Spaccapietra, S., Zimányi, E.: Conceptual modeling for traditional and spatio-
temporal applications—the MADS approach. Berlin Heidelberg: Springer Verlag (2006)
19. Theodoulidis, B., Alexakis, P., Loucopoulos, P.: Verification and validation of temporal busi-
ness rules. In: DAISD. pp. 1–15 (1992)