AN ADVANCED TOOL FOR MANAGING FUZZY COMPLEX TEMPORAL INFORMATION

Jan Zizka et al. (Eds) : CCSIT, SIPP, AISC, CMCA, SEAS, CSITEC, DaKM, PDCTA, NeCoM - 2016
pp. 97–116, 2016. © CS & IT-CSCP 2016 DOI : 10.5121/csit.2016.60110
AN ADVANCED TOOL FOR MANAGING
FUZZY COMPLEX TEMPORAL
INFORMATION
Aymen Gammoudi1,2
, Allel Hadjali2
and Boutheina Ben Yaghlane3
1
LARODEC/ISGT, 92, Boulevard 9 Avril 1938, 1007 Tunis, Tunisia
aymen.gammoudi@ensma.fr
2
LIAS/ENSMA, 1 Avenue Clement Ader, 86960 Futuroscope Cedex, France
allel.hadjali@ensma.fr
3
IHECT, University of Carthage,
IHEC-Carthage Presidence, 2016 Tunis, Tunisia
boutheina.yaghlane@ihec.rnu.tn
ABSTRACT
Many real-life applications need to handle and manage time pieces of information. Allen
temporal relations are one of the most used and known formalisms for modelling and handling
temporal data. This paper discusses a novel idea to introduce some kind of flexibility in defining
such relations between two fuzzy time intervals. The key concept of this approach is a fuzzy
tolerance relation conveniently modelled. Tolerant Allen temporal relations are then defined
using the dilated and the eroded intervals of the initial fuzzy time intervals. By leveraging some
particular fuzzy indices to compare two fuzzy time intervals, this extension of Allen relations is
integrated in the Fuzz-TIME system developed in our previous works.
KEYWORDS
Temporal Relations, Fuzzy Time Intervals, Tolerance Relation, Dilation, Erosion, Temporal
Databases.
1. INTRODUCTION
Among advanced databases we find in real world applications, temporal databases which manage
temporal aspects (time, date ...) of the data they contain are often encountered. In such
applications, temporal information is often perceived or expressed in an imprecise/fuzzy manner
[1]. For example, periods of global revolutions are characterized by beginnings and endings
naturally gradual and ill-defined (such as "well after early 20" or "to late 30"). Unfortunately, and
to the best of our knowledge, there is not much work devoted to querying/handling
fuzzy/imprecise information in a temporal databases context. On the contrary, in Artificial
Intelligence field, several works exist to represent and handle imprecise or uncertain information
in temporal reasoning (see for instance, Dubois et al. [2], Schockaert S. and De Cock [3],
Badaloni and Giacomin [4]).

98 Computer Science & Information Technology (CS & IT)
As mentioned above, only very few studies have considered the issue of modeling and handling
flexible queries over regular/fuzzy temporal databases. In [5] [6], an approach that integrates
bipolar classifications to determine the degree of satisfaction of records, is proposed. It relies on
using both positive and negative imprecise and possibly temporal preferences. But this approach
is still unable to model complex temporal relationships and cannot be applied in historical
temporal databases (for instance, the user may request one time period but reject a part of this
period, when specifying the valid time constraint in the query). Deng et al. [7] have proposed a
temporal extension to an extended ERT model to handle fuzzy numbers. They have specified a
fuzzy temporal query which is an extension from the TQuel language and they have introduced
the concepts of fuzzy temporal in specification expressions, selection, join and projection.
Tudorie et al. [8] have proposed a fuzzy model for vague temporal terms and their implication in
queries' evaluation. Unfortunately, this approach does not allow to model a large class of
temporal terms (such as: just after and much before). In [9], Galindo and Medina have proposed
an extension of temporal fuzzy comparators and have introduced the notion of dates in Relational
Databases (RDB) by adding two extra precise attributes on dates (VST, VET). Recently, in [10]
we have proposed an extension, named TSQLf, of SQLf language [11] by adding the time
dimension. TSQLf language allows for expressing user queries involving fuzzy criteria on time. It
is founded on the fuzzy extension of Allen temporal relations already proposed in [2].
Unfortunately, all the above approaches consider (fuzzy) temporal relations only between regular
time intervals (i.e., their lower and upper bounds are crisp instants). While in real world
applications, time intervals are often described by ill-defined bounds to better capture the
vagueness inherent to the available pieces of time information.
This paper is a step towards dealing with that issue. It proposes an extension of Allen temporal
relations to compare fuzzy time intervals. This extension relies on a particular tolerance relation
that allows associating a fuzzy time interval with two nested fuzzy time intervals (i.e., the dilated
and the eroded intervals). Based on these two nested intervals, in order to introduce some softness
in comparing fuzzy temporal entities, tolerant Allen relations are defined. Particular fuzzy indices
are used for the purpose of tolerant Allen relations computation.
The paper is structured as follows. In Section 2, we provide some background necessary to the
reading of the paper and a critical related work. In section 3, tolerant Allen relations modeling
and their handling are explicitly discussed. Then in section 4, we describe how we have
integrated this extension in our FuzzFuzzFuzzFuzz----TIMETIMETIMETIME system. Section 5 concludes the paper and sketches
some lines for future work.
2. BACKGROUND AND RELATED WORK
The purpose of this section is manifold. We begin by recalling Allen temporal relation, and then
we recall some fuzzy comparators of interest. Finally, we present the dilation and erosion
operations on fuzzy sets. This section is mainly browsed from [2].
2.1. Allen Temporal Relations
Allen [12] has proposed a set of mutually exclusive primitive relations that can be applied
between two temporal intervals. These relationships between events are usually denoted by
before (≺), after (≻), meets (m), met by (mi), overlaps (o), overlapped by (oi), during (d),

Computer Science & Information Technology (CS & IT) 99
contains (di), starts (s), started by (si), finishes (f), finished by (fi), and (≡). Their meaning is
illustrated in Table 1 (where A = [a, a'] and B = [b, b'] are two time intervals with a and a'
(respectively b and b') represent the two bounds of A (respectively B), with a < a'.
2.2. Fuzzy comparators
In this section, we recall two fuzzy comparators expressed in terms of difference of values. Such
comparators capture approximate equalities and graded inequalities.
Approximate Equalities and Graded Inequalities
An approximate equality between two values, here representing dates, modeled by a fuzzy
relation E with membership function µE (E stands for "equal"), can be defined in terms of a
distance such as the absolute value of the difference. Namely,
µE(x, y) = µL(|x - y|)
For simplicity, fuzzy sets and fuzzy relations are assumed to be defined on the real line.
Approximate equality can be represented by
∀ x, y ∈ ℝ
µ୉
ሺx, yሻ = µ୐
ሺ|x − y|ሻ = max൭0, min ቆ1,
δ + ε − |x − y|
ε
ቇ൱
=
‫ە‬
‫۔‬
‫ۓ‬
1 if |x − y| ≤ δ
0 if |x − y| > δ + ε
δ + ε − |x − y|
ε
otherwise
Table 1. Allen Relations.

where ρ and ϵ are respectively positive and strictly positive parameters which affect the
approximate equality. With the following intended meaning: the possible values of the difference
a - b are restricted by the fuzzy set L =
Similarly, a more or less strong inequality
"greater"), of the form
In the following, we take
∀ x, y ∈ Թ
µୋ
ሺx, yሻ ൌ µ୏
ሺx െ yሻ ൌ max
ൌ
‫ە‬
‫۔‬
‫ۓ‬
x
We assume ߩ > 0, i.e. G more demanding than the idea of "strictly greater" or "clearly greater". K
= ሺλ, λ ൅ ρ, ൅∞, ൅∞ሻ is a fuzzy interval which gathers all the values equal to or greater than a
value fuzzily located between λ
membership function. See Figure 1. According to the values of parameter
which indicates how much larger than
“moderately”, “much”. In a given context G(0) stands for ‘>’.
Figure 1. Modeling "approximate equality" and "graded strict inequality".
2.3. Dilation and Erosion Operations
Let us consider a fuzzy set A representing a time interval, and
E(L). A can be associated with a nested pair of fuzzy sets when using the parameterized relation
E(L) as a tolerance relation [2]. Indeed,
1
A = (A, B, a, b) stands for the trapezoi
(resp. [A-a, B+b]) is the core (resp. support
Computer Science & Information Technology (CS & IT)
are respectively positive and strictly positive parameters which affect the
are restricted by the fuzzy set L = ሺ−δ, δ, ε, εሻ 1
. In particular a E(0) b means
more or less strong inequality can be modeled by a fuzzy relation G (G stands for
µG (x, y) = µK (x – y)
൭0, min൬1,
‫ݔ‬ െ ‫ݕ‬ െ ߣ
ρ
൰൱
‫ە‬
‫ۓ‬
1 if x ൐ ‫ݕ‬ ൅ ߣ ൅ ߩ
0 if x ൑ y ൅ λ
x െ y െ λ
ρ
otherwise
> 0, i.e. G more demanding than the idea of "strictly greater" or "clearly greater". K
is a fuzzy interval which gathers all the values equal to or greater than a
and ρ. K is thus a fuzzy set of positive values with an increasing
membership function. See Figure 1. According to the values of parameter λ ൅ ρ
which indicates how much larger than b is a, may be linguistically labelled by “Slightly”,
“moderately”, “much”. In a given context G(0) stands for ‘>’.
Figure 1. Modeling "approximate equality" and "graded strict inequality".
ion and Erosion Operations
representing a time interval, and an approximate equality
can be associated with a nested pair of fuzzy sets when using the parameterized relation
E(L) as a tolerance relation [2]. Indeed,
stands for the trapezoidal membership function (t.m.f.) of the fuzzy set A where
a, B+b]) is the core (resp. support).
are respectively positive and strictly positive parameters which affect the
means a = b.
can be modeled by a fuzzy relation G (G stands for
> 0, i.e. G more demanding than the idea of "strictly greater" or "clearly greater". K
is a fuzzy interval which gathers all the values equal to or greater than a
. K is thus a fuzzy set of positive values with an increasing
the modality,
, may be linguistically labelled by “Slightly”,
an approximate equality relation
can be associated with a nested pair of fuzzy sets when using the parameterized relation
dal membership function (t.m.f.) of the fuzzy set A where [A, B]

one can build a fuzzy set of temporal instants close to A such that A ⊆ AL
. This is the
dilation operation,
one can build a fuzzy set of temporal instants close to A such that AL ⊆ A. This is the
erosion operation.
2.2.1. Dilation operation
Dilating the fuzzy set of temporal instants A by L will provide a fuzzy set AL
defined by
µ୅ై ሺrሻ = supୱ min ሺµ୉ሾ୐ሿ
ሺs, rሻ, µ୅
ሺsሻሻ (1)
= supୱ min ሺµ୐
ሺr − sሻ, µ୅
ሺsሻሻ (2)
= µ୅⨁୐
ሺrሻ (3)
Hence,
A୐
= A ⨁ L
where ⨁ is the addition operation extended to fuzzy sets [13]. AL
gathers the elements of A and
the elements outside A which are somewhat close to an element in A. See Figure 2.
One can easily check that the fuzzy set of temporal instants AL
is less restrictive than A, but still
semantically close to A. Thus, AL
can be viewed as a relaxed variant of A. In terms of t.m.f., if A
= (a, a’, α, α’)2
and L = ሺ−δ, δ, ϵ, ϵሻ then AL
= ሺa − δ, a′
+ δ, α + ϵ, α′
+ ϵ), see Figure 2.
Example:
If A = (15, 19, 2, 1) and L = (-1, 1, 0.5, 0.5) then
AL
= (14, 20, 2.5, 1.5)
Figure 2. Dilated and eroded time intervals of a fuzzy set of temporal instants A.
2
with [a, a'] (resp. [a-α, a'+α']) represents the core (resp. support) of A.

2.2.2. Erosion operation
Let L ⨁ X = A be an equation where X is the unknown variable. Solving this equation has
extensively been discussed in [14]. It has been demonstrated that the greatest solution of this
equation is given by Xഥ = A ⊖ (-L) = A ⊖ L since L = -L and where ⊖ is the extended
Minkowski subtraction defined by [11]:
µ୅⊝୐
ሺrሻ = infୱ ቀµ୐
ሺr − sሻ ⇒ॻ µ୅
ሺsሻቁ (4)
where ॻ a t-norm, and ⇒ॻ is the R-implication induced by ॻ and defined by ⇒ॻ ሺu, vሻ =
sup{λ ∈ ሾ0, 1ሿ, ॻ ሺu, λሻ ≤ v} for u, v ∈ ሾ0, 1ሿ. We make use of the same t-norm ॻ ሺ= minሻ as in
the dilation operation which implies that ⇒ॻ is the so-colled Goሷdel implication.
Let ሺEሾLሿሻ୰ = {s, µ୉ሾ୐ሿ
ሺs, rሻ > 0} be the set of elements that are close to r in the sense of E[L].
Then, the above expression can be interpreted as the degree of inclusion of ሺEሾZሿሻ୰ in A. This
means that r belongs to A ⊖ L all the elements s that are close to r are A. Hence, the inclusion A
⊖ L ⊆ A holds. This operation is very useful in natural language to intensify the meaning of
vague terms. Now, eroding the fuzzy set A by L results in the fuzzy set AL defined by
AL= A ⊖ L.
The fuzzy set AL is more precise than the original fuzzy set A but it still remains not too far from
A semantically speaking. If A = ሺa, a′
, α, α′
ሻ and L = ሺ−δ, δ, ϵ, ϵሻ then
A୐ = A ⊝ L = ሺa + δ, a′
− δ, α − ϵ, α′
− ϵሻ provide that α ≥ ϵ and α′ ≥ ϵ. Figure 2 illustrates
this operation.
Example:
If A = (15, 19, 2, 1) and L = (-1, 1, 0.5, 0.5) then
AL = (16, 18, 1.5, 0.5)
In the crisp case, A ⊝ L = ൣa, a′
൧ ⊝ ൣ−δ, δ′
൧ = ሾa + δ, a′
− δ′ሿ (while A ⨁ Z = ሾa − δ, a′
+ δ′ሿ).
One can easily check that the following proposition holds:
Proposition 1. Using the t.m.f. of AL
and AL given above, we have:
• ሺA୐
ሻ୐ = ሺA୐ሻ୐
= A
• ሺA୐
ሻ୐
= A ⨁ 2L
• ሺA୐ሻ୐ = A ⊖ 2L
2.4. Related Work
Temporal information is often perceived or expressed in a vague and imprecise manner. Here we
discuss some works related to the modeling and handling of imperfections in time both in
Artificial Intelligence and Databases fields.
The treatment of imprecise or uncertain information in temporal reasoning has been addressed for
a longtime. Dubois and Prade in [15] discuss the approximate reasoning on fuzzy dates and fuzzy

intervals in the framework of possibility theory. Guesgen et al. [16] propose fuzzy Allen relations
viewed as fuzzy sets of ordinary Allen relationship taking into account a neighborhood structure.
Fuzzy sets, which play a key role in the modeling of flexible constraints, have also been used in
different approaches based on constraints for temporal reasoning. Qian and Lu [17] have studied
several propagation strategies for the treatment of fuzzy rules temporal networks. Barro et al. [18]
have proposed a generalization based on fuzzy sets of time constraint and used possibility
measures to verify the consistency's degree of a fuzzy temporal constraint network. Godo and
Vila [19] have defined a temporal logic approximated based on the integration of fuzzy
constraints in a logical language. The inference system is based on specific rules treating fuzzy
constraints and proposals degrees of certainty. Dubois et al. [20] have proposed a possibilistic
temporal logic that a formula which is associated with a fuzzy set of dates when the formula is
more or less certainly true.
On the other hand, only few works have been proposed for dealing with imperfect data in
databases. Billiet et al. [5] have proposed an approach that integrates bipolar classifications to
determine the degree of satisfaction of records by using both positive and negative imprecise and
possibly temporal preferences. But this approach is still unable to model complex temporal
relationships and not applied in historical temporal databases (for instance, the user may request
one time period but reject a part of this period, when specifying the valid time constraint in the
query). Tudorie et al. [8] have proposed a fuzzy model for vague temporal terms and their
implication in queries' evaluation. Unfortunately, this approach does not allow to model a large
number of temporal terms (such as: just after and much before). Galindo and Medina [9] have
proposed an extension of temporal fuzzy comparators and have introduced the notion of dates in
Relational Databases (RDB) by adding two extra precise attributes on dates (VST, VET). The
most disadvantage of this approach that cannot support some sophisticated queries that need a
step of reasoning before processing. However, some comparators might not be in full agreement
with the intuitive semantics underlying the notion of the temporal relations that refer to.
However, in [10] we have proposed an extension, named TSQLf, of SQLf language [11] by the
addition of the dimension time. TSQLf language allows for expressing user queries involving
fuzzy criteria on time. It is founded on the fuzzy extension of Allen temporal relations already
proposed in [2]. Recently, in [21] we have proposed a first step to introduce some flexibility in
defining such relations between two fuzzy time intervals. This idea presents an extension of Allen
temporal relations based on a particular tolerance relation that allows associating a fuzzy time
interval with two nested intervals (i.e., the dilated and the eroded intervals).
Unfortunately, all the above approaches consider (fuzzy) temporal relations only between regular
time intervals (i.e., their lower and upper bounds are crisp instants). While in real world
applications, time intervals are often described by ill-defined bounds to better capture the
vagueness inherent to the available pieces of time information. In this context (i.e., bounds of
temporal intervals are ill-defined), not much work exist in the literature. Except, the works done
by Nagypal and Motik [22], Ohlbach [23] and Schockaert et al. [3]. Nagypal and Motik [22] have
defined a temporal model based on fuzzy sets. This model extends Allen relations with fuzzy
time intervals (ITFs). An ITF means a temporal interval with bounds defined in an imprecise way
(for example, "the period from the late 20s to the early 30s" is an ITF with the following
semantics (1928, 1933, 2, 2)). Nagypal and Motik have introduced a set of auxiliary operators on
intervals such as, for example, the operator taking an interval I and built intervals containing all

instants which are before initial time I. Then, fuzzy counterparts of these operators have been
defined on the ITFs. Extended Allen relations with ITFs were introduced using the fuzzy
operators. Note that the composition of these relations was not discussed by the authors.
Schockaert et al. [3] also proposed a generalization of Allen relations with ITFs. This
generalization allows handling classical relations between imprecise events (such as, "Roosevelt
died before the start of the Cold War"), and also imprecise nature relations (such as, "Roosevelt
died just before the start of the Cold War "). The key notion used in this approach is the concept
of fuzzy orders on time (as, for example, the fuzzy order that expresses how a moment a is much
smaller than a moment b). These orders represented by parameterized fuzzy relations are applied
to the gradual bounds ITFs to define, for example, the degree bb<<
(A, B) (expressing how the
beginning of an ITF A, defined in an imprecise manner, is before the end of an ITF B defined in
an imprecise way as well). In [4], Badaloni and Giacomin have developed a fuzzy extension of
the algebra of intervals, named IAfuz
where degrees of preference (belonging to [0, 1]) are
attached to each atomic relation between two classical time intervals.
3. TOLERANT ALLEN RELATIONS
The purpose of this section is twofold. We start by modeling tolerant Allen relations based on
dilation and erosion operations. Then, we present the comparing indices of two fuzzy intervals to
compute to what extent a tolerant Allen relation is satisfied.
3.1. Modeling
Using the dilatation and erosion operations defined above, we can provide the basis for defining a
tolerance-based extension of Allen relations. For instance, the tolerance-based temporal relation
corresponding to meet Allen relation writes:
A toler-meets(L) B would correspond to AL before BL and AL
overlaps BL
The statement A toler-meets(L) B means that the temporal relation between the two (crisp/fuzzy)
temporal intervals A and B is perceived as a variant of meet relation thanks to some tolerance
expressed by the indicator L. This is a human perception which is often encountered in real world
applications (such as in managing historical temporal data, planning and scheduling, natural
language understanding, etc.).
The toler-meet relation gathers a class of Allen relations (i.e., before and overlaps) applied on the
eroded and dilated time intervals corresponding to the original time intervals A and B. This boils
down to compute the traditional Allen relations, before and overlaps, on fuzzy temporal intervals
AL
and AL (since AL
and AL are fuzzy sets as shown in the previous subsection).
Let A = ሺa, a′
, α, α′
ሻ be a fuzzy time interval with a෥ = ሺa, a, α, 0ሻ and a෥′ = ሺa′, a′, 0, α′ሻ the two
fuzzy bounds of validity of A. One can write A = ሺa෤, a, a′
, a෤′ሻ. Under these forms:
AL
writes ሺa − δ, a′
+ δ, a෤ሺ୐ሻ
, a෤′ሺ୐ሻ
ሻ where a෤ሺ୐ሻ
= ሺa − δ, a − δ, α + ϵ, 0ሻ and
a෤′ሺ୐ሻ
= ሺa′
+ δ, a′
+ δ, 0, α′ + ϵሻ.
AL writes ሺa + δ, a′
− δ, a෤ሺ୐ሻ, a′෩
ሺ୐ሻ
ሻ where a෤ሺ୐ሻ = ሺa + δ, a + δ, α − ϵ, 0ሻ and
a෤′ሺ୐ሻ = ሺa′
− δ, a′
− δ, 0, α′ − ϵሻ.

Example:
Let A a fuzzy temporal interval representing the period from the early 20 until the end of 20. It is
easier to see that A can writes A = ሺܽ෤, ܽ෤′
, ߙ, ߙ′ሻ.
With ܽ෤ = (1920, 1920, 3, 0) and ܽ෤′ = (1930, 1930, 0, 2).
Given that α = 3, α’ = 2 and L = (-3, 3, 1, 1).
Then
AL
= ሺ 1917, 1933, ܽ෤ሺ௅ሻ
, ܽ෤′ሺ௅ሻ
ሻ where ܽ෤ሺ௅ሻ
= (1917, 1917, 4, 0) and ܽ෤′ሺ௅ሻ
= (1933, 1933, 0, 3).
And
AL = ሺ1923, 1927, ܽ෤ሺ௅ሻ, ܽ෤′ሺ௅ሻሻ where ܽ෤ሺ௅ሻ = (1923, 1923, 2, 0) and ܽ෤′ሺ௅ሻ = (1927, 1927, 0, 1).
In a similar way, the tolerant counterparts of all Allen relations write:
• A toler-meets (L) B as AL before BL and AL
overlaps BL
.
• A toler-before(L) B as AL
toler-meets(L) BL
.
• A toler-overlaps (L) B as AL toler-meets BL.
• A toler-during (L) B as AL
toler-equals BL.
• A toler-starts(L) B as AL during BL
and AL
overlaps BL.
• A toler-finishes (L) B as AL
overlapped_by BL ˄ AL during BL
.
• A toler-equals (L) B as AL
contains BL ˄ AL during BL
.
3.2. Computation
In this section, firstly, we recall the comparing indices of two fuzzy intervals proposed by Dubois
and Prade in [24].
Let two fuzzy time intervals M and N expressed by quadruples of the form (m1, m2, α1, α2) and
(n1, n2, ß1, ß2) respectively. Basically, there are four indices to interpret how M is greater than N
(see Figure 3).
dሺM ≻ Nሻ = inf୶,୷{max ቀ1 − µ୑
ሺxሻ, 1 − µ୒
ሺyሻቁ : x ≤ y} (5)
dሺM ≻ା
Nሻ = inf୶sup୷{maxቀ1 − µ୑
ሺxሻ, µ୒
ሺyሻቁ : x ≥ y} (6)
dሺM ≻ି
Nሻ = sup୶inf୷{maxቀµ୑
ሺxሻ, 1 − µ୒
ሺyሻቁ : x < ‫}ݕ‬ (7)
dሺM ≽ Nሻ = sup୶,୷{minቀµ୑
ሺxሻ, µ୒
ሺyሻቁ : x ≥ y} (8)

Figure 3. Possibilistic indices of fuzzy intervals comparisons.
Equation (5) expresses the certainty that x is greater than N, knowing that x is M. This means that
M is necessarily greater than N. It can be expressed on the basis of a necessity degree of the
proposal that M is strictly greater than N as follows:
ℵ୑ሺሿN, +∞ሻሻ = 1 − sup୶ஸ୷min ሺµ୑
ሺxሻ, µ୒
ሺyሻሻ (9)
Equation (9) refers to the degree of inclusion of the fuzzy set M in ]N, + ∞). Knowing that;
ℵ୑ሺሿN, +∞ሻሻ = Ness ሺx > ܰ | ‫ݔ‬ ݅‫ݏ‬ ‫ܯ‬ሻ
Equation (6) expresses the certainty that x is greater or equal to N, knowing that x is M. It can be
expressed on the basis of a necessity degree of the proposal that M is greater than or equal to N as
follows:
ℵ୑ሺሾN, +∞ሻሻ = inf୶sup୷ ∶୷ ஸ୶max ሺ1 − µ୑
ሺxሻ, µ୒
ሺyሻሻ (10)
Equation (10) refers to the degree of inclusion of the fuzzy set M in [N, +∞). Knowing that
ℵ୑ሺሾN, +∞ሻሻ = Ness ሺx ≥ N | x is Mሻ
Equation (7) expresses the possibility that x is greater than N, knowing that x is M. It can be
expressed on the basis of a possibility degree of the proposal that M is strictly greater than N as
follows:
Π୑ሺሿN, +∞ሻሻ = sup୶inf୷ ∶୷ ஹ ୶min ሺμ୑ሺxሻ, 1 − μ୒ሺyሻሻ (11)
Equation (11) refers to the degree of nonemptiness of the fuzzy set M ∩ ሿN, +∞ሻ of numbers
strictly greater than N, given that they are restricted by M. Knowing that:
∏୑ሺሿN, +∞ሻሻ = Poss ሺx > ܰ | ‫ݔ‬ ݅‫ݏ‬ ‫ܯ‬
Equation (8) expresses the possibility that x is greater or equal to N, knowing that x is M. It can
be expressed on the basis of a possibility degree of the proposal that M is greater than or equal to
N as follows:

∏୑ሺሾN, +∞ሻሻ = sup୶,୷ ∶୶ ஹ ୷min ሺµ୑
ሺxሻ, µ୒
ሺyሻሻ (12)
Equation (12) refers to the degree of nonemptiness of the fuzzy set M ∩ ሾN, +∞ሻ of numbers
greater than or equal to N, given that they are restricted by M. Knowing that:
∏୑ሺሾN, +∞ሻሻ = Poss ሺx ≥ N | x is Mሻ.
Given these degrees of comparison between two fuzzy time intervals, we can use equation (5)
redefined in terms of a degree of necessity by the equation (9) to assess the extent to which a
fuzzy time interval A is greater than another fuzzy time interval B (see Figure 4), denoted by
dሺA ≻ Bሻ (with A = (a1, a2, α1, α2) and B = (b1, b2, ß1, ß2)) as follows;
dሺA ≻ Bሻ = 1 − sup୶ஸ୷min ሺµ୅
ሺxሻ, µ୆
ሺyሻሻ (13)
= ቐ
1 if aଵ − αଵ ≥ bଶ + βଶ
1 − ρ if aଵ − αଵ < bଶ + βଶ
and aଵ > bଶ
0 otherwise
with ρ =
୶ିሺୟభିαభሻ
αభ
and x =
ୠమαభା ୟభ
ଵା αభ
To illustrate formula (13), let us consider the following example (where A = [25/10/2015,
28/10/2015, 1, 1] that expresses time around October 25 and October 28):
Case 1: B = [19/10/2015, 20/10/2015, 2, 2]
aଵ − αଵ > bଶ + βଶ
24/07/2015 > 20/07/2015 then
dሺA ≻ Bሻ = 1
Case 2: B = [19/10/2015, 24/10/2015, 2, 2]
aଵ − αଵ < bଶ + βଶ
24/07/2015 < 26/07/2015 and
aଵ > bଶ 25/07/2015 > 24/07/2015 then
dሺA ≻ Bሻ = 1 − ρ = 1 −
ୠమିୟభାଵାαభ
ଵାαభ
= 0,5
Case 3: B = [19/10/2015, 26/10/2015, 2, 2]
aଵ − αଵ < bଶ + βଶ
24/07/2015<27/07/2015 and
aଵ < bଶ 25/10/2015 < 26/10/2015 then
dሺA ≻ Bሻ = 0

Figure 4. Comparison index dሺA ≻ Bሻ.
Using the formula (13), we define in Table 2 a way to calculate the degree of tolerant Allen
relations. We have used this formula because we need to be sure that an interval A is greater than
another B. This means that A is necessarily greater than B.
Table 2. Tolerant Allen Relations (where dሺb෨୐ ≻ a෤′୐ሻ = 1 − sup୶ஸ୷min ሺµୠ෩ై
ሺxሻ, µୟ෤′ై
ሺyሻሻ).

4. FUZZ-TIME SYSTEM
4.1. Architecture Overview
Here we give an overview of Fuzz-TIME system. It includes in fact two main steps and each
step contains a set of modules.
Step 1 begins with a temporal gradual query proposed by the user through a GUI of the Fuzz-
TIME system. The latter requests the user to define a validity interval for each fuzzy temporal
specification. Then, TSQLf query is generated and sent to a main interpretation module. This
module is composed by a set of submodules; each presents an alternative for managing TSQLf
queries. The result of this module corresponds to an evaluation of an SQL-like query.
The second step firstly proceeds to pass the query by reasoning module. The latter uses the
inference machinery of fuzzy Allen relations. Then the request goes through the data
management system to select rows that meet the required temporal criteria. Finally, the selected
lines pass through the module of our system that calculates the degree of satisfaction of each line
with the selection criteria. Then, the selected lines with their degrees are displayed to the user.
4.2. Modules Description
In [10] we have proposed an extension, named TSQLf, of SQLf language [11] by adding the time
dimension. TSQLf language allows for expressing user queries involving fuzzy criteria on time.
It is founded on the fuzzy extension of temporal Allen relations already proposed in [2]. We have
implemented this language and developed a first version of the Fuzz-TIME3
system.
In this preliminary version of Fuzz-TIME, queries involving fuzzy temporal criteria can be
handled where the temporal relations can be defined in a fuzzy way but time are defined only in
terms of regular (crisp) intervals. While in real-life applications, time intervals are often
described by ill-defined bounds to better capture the vagueness inherent to the available pieces of
time information. For this reason, in this work, we have introduced tolerant temporal relations to
deal with this issue. Also, we have endowed our Fuzz-TIME Tool with the capabilities for
handling such novel temporal relations.
First, we present the architecture of Fuzz-TIME system. Figure 5 depicts the different modules
necessary to processing an TSQLf query. In the Interface module, the user enters a gradual
temporal query using a graphical interface, this latter gives him/her in the first place the
possibility of choosing attributes, tables and built fuzzy temporal conditions, and second the
system asks the user to identify the validity interval and the tolerance interval for each fuzzy
temporal condition (i.e., the fuzzy parameter L).
- If the user provides a validity interval with crisp bounds, then Fuzzy Allen Relation
module (denoted by FAR) is triggered. More details about this module can be found in
[10]. Note that in the case where the TSQLf query requires a phase of reasoning, the
system calls the module Reasoning Step. This module leverages the inference machinery
of fuzzy Allen relations developed in the first version.
3
FuzzFuzzFuzzFuzz----TIMETIMETIMETIME: Fuzzy Temporal Information Managing and Exploitation.

- If the user gives a validity interval with fuzzy bounds, then the system automatically
computes the eroded and dilated intervals. Thereafter, the system passes the request to
the Query Interpretation module which transforms the request into a crisp query using
the Tolerant Allen Relations module (denoted by TAR) using the principle of dilation and
erosion operations. The result of this module is an TSQLf query to be sent to the
management system database in order to select the attributes that meet the fuzzy temporal
query criteria.
Figure 5. Architecture of Fuzz-TIME system.
The last module, Satisfaction Degree Computation, ensures the calculation of satisfaction degree
of the TSQLf query at hand. The returned results are attached with a degree of satisfaction and
then displayed on the user interface.

4.3. An Illustrative Example
To better explain our proposal we present below an example from Archaeology field. Consider
the Archaeology table (see Table 3) presenting the material remains from prehistoric times. The
table schema is Archaeology (Code_Ar#, Name_Ar, Location, Date_Discovery, VST_Dc,
VET_Dc, Date_Dated, VST_Dd, VET_Dd). Where VST_Dc means the start validity date of Date_
Discovery, VET_Dc means the end validity date of Date_ Discovery, VST_Dd means the start
validity date of Date_Dated and VET_Dd means the end validity date of Date_Dated.
Table 3. The Archaeology table.
4.4. Demonstrative Scenarios
We will demonstrate the functioning of our FuzzFuzzFuzzFuzz----TIMTIMTIMTIMEEEE system by presenting some particular
queries over archaeological databases.
• Q1: Show archaeological discoveries that took place during the 3rd quarter of 2000.
Here the user asked a query with a fuzzy temporal condition (3rd quarter) but we can easily
interpret the validity interval of the fuzzy temporal condition corresponding to [01/07/2000,
30/09/2000]. So the solution is a simple SQL query.
Select *
From Archaeology A
Where A.VST > 01/07/2000 and A.VET < 30/09/2000;
• Q2: Show archaeological discoveries that took place just after the Second World War.
The user indicated in this query a fuzzy temporal constraint (Second World War). So we must
proceed with the generation of a TSQLf request. According to the principle of TSQLf query, the

system must ask the user to specify a valid interval for fuzzy temporal constraint proposed by the
user. Indeed two cases are possible here:
i) the user can define a validity interval with two crisp bounds [VST, VET]. For
example [01/09/1939, 02/09/1945]
So here we use the first principle of the TSQLf approach and the request will be
Select *
From Archaeology A
Where A.VET > 02/09/1949;
ii) the user can define a validity interval with two fuzzy bounds [VST, VET]. For
example [Before the end of 1939, Just before end of 1945]
So the bounds of validity interval defined by the user are fuzzy, consequently
TSQLf query should use the principle of Tolerant Allen Relations; the system
asks the user to define a fuzzy set A = (a, a’, α, α’) for the validity fuzzy interval
and a fuzzy set L = (-δ, δ, ε, ε). For example: A = (20/09/1939, 10/09/1945, 7, 5)
and L = (-3, 3, 2, 1). Then the system generates automatically two intervals the
eroded and dilated ones. Thereafter it generates a TSQLf query using Tolerant
Allen Relations proposed in section 3. 2.
4.5. Implementation and Interface
We present in this section some user-friendly interfaces that help to make gradual temporal
queries based on tolerant Allen relations. It can help users for expressing temporal terms in a
fuzzy way. The tool we have developed acts as an JAVA interface with the Oracle DBMS and
generates TSQLf queries directly executable by calls to functions and PL/SQL bloc stored. The
interface is connected to the database so as to store the tables in the field study that incorporates
fuzzy temporal aspects. In this way, it is possible to add for each table, which contains a
crisp/fuzzy temporal attribute, two specific temporal attributes (VST, VET).
Our first approach proposed in [10] is still functional in our Fuzz-TIME system, Figure 6 shows
the possibility that a user can enter a query with a temporal fuzzy constraint then he can define
the validity interval with two crisp dates.
The extension integrated in Fuzz-TIME tool allows, first, to accept the definition of fuzzy
temporal condition (Figure 7), and second, the definition of validity interval and tolerance
interval for each fuzzy temporal condition. Here the user can introduce fuzzy bounds for each
validity intervals (Figure 7).

Figure 6. Definition of fuzzy temporal conditions with crisp bounds.
Figure 7. Definition of fuzzy temporal conditions.
After the step above, the tool informs the user that the bounds introduced are fuzzy so it is
necessary to define the fuzzy set
fuzzy set L. The aim is to generate automatically the pair intervals (dilated and eroded ones).
Then, the generated TQSLf query is submitted to the database to select rows that match its fuzzy
temporal criteria. Finally, a degree of satisfaction is calculated and assigned
(Figure 8 and Figure 9).
Figure 6. Definition of fuzzy temporal conditions with crisp bounds.
Figure 7. Definition of fuzzy temporal conditions.
fuzzy set A corresponding to a fuzzy temporal condition as well as the
aim is to generate automatically the pair intervals (dilated and eroded ones).
temporal criteria. Finally, a degree of satisfaction is calculated and assigned to each selected line
113
corresponding to a fuzzy temporal condition as well as the
aim is to generate automatically the pair intervals (dilated and eroded ones).
to each selected line

Figure 8. Result of an TSQLf query.
Figure 9. Result of an TSQLf query

5. CONCLUSION
In this paper, we have defined the principle of tolerance-based of Allen temporal relations to
manage time intervals with fuzzy bounds. The key notion of this extension is the dilation and
erosion operations defined on fuzzy time intervals. Then we have defined a way for computing
such introduced relations by leveraging fuzzy indices comparison. A new version of our Fuzz-
TIME System is developed. It allows handling temporal queries where time can be expressed in
terms of fuzzy intervals.
As for future immediate work, we first plan to establish the complete set of composition rules of
the tolerant Allen relations for the purpose of reasoning and inference. Second, we incorporate
such reasoning in our Fuzz-TIME System. Another line of future research is to investigate the
issue of uncertainty in temporal relations in the spirit of [25].
REFERENCES
[1] C. Billiet & G. De Tré, (2015) “The Role of Computational Intelligence in Temporal Information
Retrieval: A Survey of Imperfect Time in Information Systems”, In Proc. of the 13th International
Conference on Flexible Query Answering Systems (FQAS’2015), Vol. 400, pp. 41-756.
[2] D. Dubois, A. Hadjali, & H. Prade, (2003) ‘’Fuzziness and uncertainty in temporal reasoning’’,
Journal of Universal Computer Science, Special Issue on Spatial and Temporal Reasoning, Vol. 9, pp.
1168-1194.
[3] S. Schockaert & M. De Cock, (2008) ‘’Temporal reasoning about fuzzy intervals’’, Artificial
Intelligence, Vol. 172, pp. 1158-1193.
[4] S. Badaloni & M. Giacomin, (2006) “The algebra IAfuz: a framework for qualitative fuzzy
temporal”, Artificial Intelligence, Vol. 170 (10), pp. 872-902.
[5] C. Billiet, J. E. Pons, T. Matthe, G. De Tré, & O. P. Capote, (2011) “Bipolar fuzzy querying of
temporal databases”, In Proc. of the 9th International Conference on Flexible Query Answering
Systems (FQAS’2011), Springer Berlin Heidelberg, Vol. 7022, pp. 60-71.
[6] F. Pons, C. Billiet, O. Pons, G. De Tré, (2014) “Aspects of dealing with imperfect data in temporal
databases”. In: Pivert, O., Zadro´zny, S. (eds.) Flexible Approaches in Data, Information and
Knowledge Management, Springer, Heidelberg, Vol. 497, pp. 189–220.
[7] L. Deng, Z. Liang, & Y. Zhang, (2008) “A fuzzy temporal model and query language for fter
databases’’, 8th International Conference on Intelligent Systems Design and Applications, Vol. 3, pp.
77-82
[8] C. Tudorie, M. Vlase, C. Nica, & D. Muntranu, (2012) “Modeling fuzzy temporal criteria in database
querying’’, Artificial Intelligence Applications, Vol. 112, pp. 1-6
[9] J. Galindo & J.M.Medina, (2001) “Ftsql2 : Fuzzy time in relational databases’’, Proc. of the 2nd
International Conf. in Fuzzy Loggic and Technology, pp. 5-7
[10] A. Gammoudi, A. Hadjali, & B B. Yaghlane, (2014) “An intelligent flexible querying approach for
temporal databases’’, Proc. of the 7th international Conference on Intelligent Systems IS’14, Vol.
322, pp. 523-534.
[11] O. Pivert & P. Bosc, (2012) “Fuzzy Preference Queries to Relational Databases”, Imperial College
Press, pp. 330
[12] J. F. Allen, (1983) “Maintaining knowledge about temporal intervals’’, Comm. of the ACM, Vol. 26,
pp. 832-843
[13] D. Dubois & H. Prade, (1988) “Possibility theory’’, Plenum Press.
[14] D. Dubois & H. Prade, (1983) “Inverse operations for fuzzy numbers’’, Proc. IFAC Symp. on Fuzzy
Info., Knowledge representation and Decision Analysis, pp. 391-395.
[15] D. Dubois & H. Prade (1989) “Processing fuzzy temporal knowledge”, IEEE Trans. On Systems,
Man and Cybernetics, Vol. 19, pp. 729-744.

[16] H. W. Guesgen, J. Hertzberg, & A. Philpott, (1994) “Towards implementing fuzzy allen relations”,
Proc. ECAI-94 Workshop on Spatial and Temporal Reasoning, Amsterdam, The Netherlands, pp. 49-
55.
[17] D.Q. Qian & Y.Z. Lu, (1989) “A strategy of problem solving in a fuzzy reasoning network”, In Fuzzy
Sets and Systems, Vol. 33, pp. 137-154.
[18] S. Barro, R. Marin, J. Mira, & A. Paton, (1994) “A model and a language for the fuzzy representation
and handling of time”, In Fuzzy Sets and Systems, Vol. 31. pp. 153-175.
[19] L. Godo & L. Vila, (1995) “Possibilistic temporal reasoning based on fuzzy temporal constraints”, In
Proc. of the 14th Inter. Joint Conf. on Artiﬁcial Intelligence (IJCAI’95), Montral, Vol. 2, pp. 1916-
1922.
[20] D. Dubois, J. Lang, & H. Prade (1991), “Timed possibilistic logic”, In Fundamentae Informaticae,
Vol. 15, pp. 211-234.
[21] A. Gammoudi, A. Hadjali, & B B. Yaghlane, (2015) “A Tolerance-Based Semantics of Temporal
Relations: First Steps”, Proc. of the 7th International Conference on Computational Collective
Intelligence Technologies and Applications ICCCI’15, Spring Vol. 9329, pp. 453-462.
[22] G. Nagypal & B. Motik, (2003) “A fuzzy model for representing uncertain, subjective, and vague
temporal knowledge in ontologies”, On The Move to Meaningful Internet Systems 2003: CoopIS,
DOA, and ODBASE, LNCS, Vol. 2888, pp. 906–923
[23] H. Ohlbach (2004), “Relations between fuzzy time intervals’’. Proc. of the 11th International
Symposium on Temporal Representation and Reasoning, pp. 44-51
[24] D. Dubois & H. Prade, (1983) “Ranking fuzzy numbers in the setting of possibility theory’’,
Informations Sciences, Vol. 30, pp. 183-224.
[25] N. El Hadj Salem, A. Hadjali, B. B. Yaghlane & A. Gammoudi, (2015) “An Evidential Approach to
Managing Temporal Uncertainty”, Proc. of the 24th French Conference on Fuzzy Logic and its
Applications, LFA’15, Poitiers, France, pp. 243-249.

AN ADVANCED TOOL FOR MANAGING FUZZY COMPLEX TEMPORAL INFORMATION

More Related Content

What's hot (17)

Similar to AN ADVANCED TOOL FOR MANAGING FUZZY COMPLEX TEMPORAL INFORMATION (20)

Recently uploaded (20)

AN ADVANCED TOOL FOR MANAGING FUZZY COMPLEX TEMPORAL INFORMATION