Expert Systems with Applications 41 (2014) 3615–3627

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

Formal translation from fuzzy EER model to fuzzy XML model

Li Yan a,⇑, Z.M. Ma b
a
School of Software, Northeastern University, Shenyang 110819, China
b
College of Information Science and Engineering, Northeastern University, Shenyang 110819, China

a r t i c l e i n f o a b s t r a c t

Keywords: XML has been the de facto standard of data representation and exchange over the Web. In addition, fuzzy
Data modeling data are inherent in the real-world applications. Although fuzzy information has been extensively inves-
Fuzzy information tigated in the context of relational database model, the classical relational database model and its fuzzy
Conceptual data model extension to date do not satisfy the need of modeling and processing complex objects with imprecision
XML model
and uncertainty on the Web. Based on fuzzy sets, this paper concentrates on fuzzy information modeling
Mapping
in the EER (enhanced entity-relationship or extended entity-relationship) model and the fuzzy XML
model. In particular, the formal approach to mapping the fuzzy EER model to the fuzzy DTD (document
type definition) model is developed.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
With the prompt development and wide applications of the
Internet, the requirement of managing Web data has attracted
much attention both from academia and industry. XML is widely
regarded as the next step in the evolution of the World Wide
Web, and has been the de facto standard. Aiming at enhancing
content on the World Wide Web, XML and related standards are
flexible and allow the easy development of applications which ex-
change data over the Web such as e-commerce (EC) and supply
chain management (SCM). But this flexibility makes it challenging
to develop an XML management system. To manage XML data, it is
necessary to integrate XML and databases (Bertino and Catania,
2001). Various databases, including relational, object-oriented,
and object-relational databases, have been applied for mapping
to and from the XML document. In addition, XML lacks sufficient
power in modeling real-world data and their complex inter-
relationships in semantics. So it is necessary to use other methods
to describe data paradigms and develop a true conceptual data
model, and then transform this model into an XML encoded for-
mat. Conceptual data modeling of XML document schema (Conrad
et al., 2000; Elmasri et al., 2005; Mani et al., 2001; Psaila, 2000;
Xiao et al., 2001) and XML Schema (Bernauer et al., 2004) have
been studied in the recent past. In (Conrad et al., 2000), for exam-
ple, UML is used for designing XML DTD (document type defini-
tion). At the same time, some data are inherently imprecise and
uncertain since their values are subjective in the real-world appli-
cations. For example, considering the values which represent the
⇑ Corresponding author.
E-mail address: yanl@swc.neu.edu.cn (L. Yan).
satisfaction degree for a film, different individuals may have very
different satisfaction degree. Information fuzziness on the Web
has been investigated in the context of EC and SCM (Petrovic
et al., 1999; Yager, 2000; Yager and Pasi, 2001). It is shown that
fuzzy sets are very useful in Web-based business intelligence.
Viewed from data modeling, fuzzy information has been exten-
sively investigated in the context of relational database model
(Buckles and Petry, 1982; Prade and Testemale, 1984; Raju and
Majumdar, 1988; Umano and Fukami, 1994). However, the classi-
cal relational database model and its fuzzy extension cannot satisfy
the need of modeling complex objects with imprecision and uncer-
tainty. The requirements of modeling complex objects and infor-
mation imprecision and uncertainty can be found in many
application domains (e.g., multimedia applications) and have chal-
lenged the current database technology (Aygun and Yazici, 2004;
Chamorro-Martínez et al., 2007). To model uncertain data and
complex-valued attributes as well as complex relationships among
objects, current efforts have concentrated on the fuzzy object-
oriented databases (Bordogna et al., 1999; Dubois et al., 1991;
George et al., 1996; Gyseghem and Caluwe, 1998; Ma et al.,
2004) and the fuzzy conceptual data models (Chen and Kerre,
1998; Galindo et al., 2004; Zvieli and Chen, 1986). In (Zvieli and
Chen, 1986), fuzzy sets are applied to some of the basic ER (en-
tity-relationship) concepts, introducing fuzzy entity sets, fuzzy
relationship sets and fuzzy attribute sets (the first level of fuzzi-
ness) as well as fuzziness in entity and relationship occurrences
(the second level of fuzziness) and in attribute values (the third
level of fuzziness). Without including graphical representations,
the fuzzy extensions of several major EER (enhanced entity-
relationship or extended entity-relationship) concepts (e.g.,
superclass/subclass, generalization/specialization, category and

0957-4174/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.eswa.2013.12.014
3616 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
the subclass with multiple superclasses) are introduced in Chen
et al. (1998). In Galindo et al., 2004, the fuzzy EER model is ex-
tended by relaxing some constraints with fuzzy quantifiers. Also
there are efforts to conceptually design the fuzzy databases using
the fuzzy conceptual data models (Ma, 2005; Ma and Shen, 2006;
Ma et al., 2001; Yazici et al., 1999; Ma et al., 2012, 2011). More re-
cently, the fuzzy object-relational databases are proposed (Cuevasa
et al., 2008) which combine both characters of fuzzy relational dat-
abases and fuzzy object-oriented databases. Ones can refer to (Ma
and Yan, 2008, 2010) for recent surveys of these fuzzy data models.
Despite fuzzy sets have been applied to model and handle
imprecise and uncertain information in databases since Zadeh
introduced the theory of fuzzy sets (Zadeh, 1965), relative little
work has been carried out in extending XML towards the represen-
tation of imprecise and uncertain concepts. Abiteboul et al.
Abiteboul et al. (2001) provide a model for XML documents and
DTDs and a representation system for XML with incomplete infor-
mation. The representations of probabilistic data in XML are pro-
posed in other previous research papers such as (Hung et al.,
2003; Nierrman and Jagadish, 2002; Senellart and Abiteboul,
2007; Van Keulen et al., 2005). Without presenting XML represen-
tation model, data fuzziness in XML document is discussed directly
according to the fuzzy relational databases in Gaurav and Alhajj
(2006), and the simple mappings from the fuzzy relational dat-
abases to fuzzy XML document are provided also. In (Oliboni and
Pozzani, 2008), an XML Schema definition is given for representing
fuzzy data, which adopts data type classification for the XML data
context. A fuzzy XML data model based on XML DTD (Document
Type Definition) is proposed in Ma and Yan (2007), in which the
mappings of the fuzzy XML DTD from the fuzzy UML data model
and to the fuzzy relational database model are discussed, respec-
tively. In (Yan, 2009), a fuzzy XML data model based on XML
Schema is developed. Ones can refer to (Ma and Yan, 2010) for re-
cent research work of soft computing in XML data management.
For the classical XML, conceptual data models are generally ap-
plied to construct XML model through mapping conceptual data
models into XML model because XML lacks sufficient power in
modeling real-world data and their complex inter-relationships
in semantics. Similarly in order to conceptually design the fuzzy
XML model, the fuzzy UML data model is applied to formally
map to the fuzzy XML DTD model in Ma and Yan (2007). It should
be pointed out that the conceptual data models for the conceptual
design of XML model are diverse and different designers may use
different conceptual data model. In the context of database model-
ing, for example, in addition to the UML data model, the EER model
is able to capture and represent rich and complex semantics at a
high abstract level and can be used for conceptual design of dat-
abases as well as XML. So in order to conceptually design the fuzzy
XML model using different fuzzy conceptual data model, in this pa-
per, we present a full-fledged fuzzy extension to the EER model and
the corresponding graphical representations. In particular, we de-
velop the formal approach to mapping the fuzzy EER model to
the fuzzy XML DTD mode. With the proposed approach, the con-
struction of the complex fuzzy XML model can start from the de-
sign of the fuzzy EER model, which is then mapped into the
fuzzy XML model automatically. Note that the fuzzy XML model
discussed in this paper only focuses on the fuzzy XML DTD. Limited
by the expressive power of XML DTD, the fuzzy XML model
mapped from the fuzzy EER model cannot support a much richer
set of structures, types and constraints for describing fuzzy data.
The remainder of this paper is organized as follows. Section 2
presents fuzzy information modeling in the fuzzy XML model
and the fuzzy EER model. In Section 3, the formal approach to map-
ping the fuzzy EER model to the fuzzy XML DTD model is devel-
oped. A small case study is given in Section 4 to show the
proposed formal translation. Section 5 concludes this paper.
2. Fuzzy information modeling in XML and EER models
The concept of fuzzy sets was originally introduced by Zadeh,
(1965). Let U be a universe of discourse and F be a fuzzy set in U.
A membership function lF: U ? [0, 1] is defined for F, where lF
(u), for each u 2 U, denotes the membership degree of u in the fuz-
zy set F. Thus, the fuzzy set F is described as follows:
F ¼ fðu1 ; lF ðu1 ÞÞ; ðu2 ; lF ðu2 ÞÞ; . . . ; ðun ; lF ðun ÞÞg
The fuzzy set F is consisted of some elements just like a conven-
tional set. But, not being the same as the conventional set, each ele-
ment in F may or may not belong to F, having a membership degree
to F which needs to be explicitly indicated. So in F, an element (say
ui) is associated with its membership degree (say lF (ui)), and they
occur together in form of (ui, lF (ui)). When the membership de-
grees that all elements in F belong to F are exactly 1, the fuzzy set
F reduces to a conventional set.
When the membership degree lF (u) above is explained to be a
measure of the possibility that a variable X has the value u, where X
takes values in U, a fuzzy value is described by a possibility distri-
bution pX (Zadeh, 1978).
pX ¼ fðu1 ; pX ðu1 ÞÞ; ðu2 ; pX ðu2 ÞÞ; . . . ; ðun ; pX ðun Þg
Here, pX (ui), ui 2 U, denotes the possibility that ui is true. Let pX be
the possibility distribution representation for the fuzzy value of a
variable X. It means that the value of X is fuzzy, and X may take
one from some possible values u1, u2, . . ., and un and each one
(say ui) taken possibly is associated with its possibility degree
(say pX (ui)).
2.1. Fuzzy XML model
Two kinds of fuzziness can be identified in XML documents (Ma
and Yan, 2007). The first one is the fuzziness in elements (we use
membership degrees associated with such elements) and the sec-
ond one is the fuzziness in attribute values of elements (we use
fuzzy sets to represent such values). For the latter, there exist
two interpretations on it (i.e., disjunctive semantics and conjunc-
tive semantics) and they may occur in child elements with or with-
out further child elements in the ancestor–descendant chain. The
basic data structure of fuzzy XML data model is the data tree
(Ma et al., 2010).
Definition. Let V be a finite set (of vertices), E 2 V  V be a set (of
edges) and ‘: E ? C be a mapping from edges to a set C of strings
called labels. The triple G = (V, E,‘) is an edge labeled directed
graph.
Based on the data tree, we introduce the definition of fuzzy XML
data tree.
Definition. Fuzzy XML data tree F is a 6-tuple, F = (V, w, ‘, s, j, d)
where
 V = {V1,. . .,Vn} is a finite set of vertices.
 w  {(Vi, Vj) | Vi, Vj 2 V}, (V, w) is a directed tree.
 ‘: V ? (L [ {null}), here L is a set of labels. For each object v 2 V
and each label » 2 L, ‘(v, ») specifies the set of objects that may
be children of v with label ».
 s ? T, T is a set of types.
 j is mapping which constrains the number of children with a
given label. j associates with each object v 2 V and each
label» 2 L, an integer-valued interval function. j (v, ») = [min, -
max], where min P 0, max P min. We use j to represent the
lower and upper bounds.
 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
d is a mapping from the set of objects v 2 V to local possibility
functions. It defines the possibility of a set of children of an
object existing given that the parent object exists.
Definition. Suppose F = (V, w, ‘, s, j, d) and f’ = (V’, w’, ‘’, s’, j’, d’) are
two fuzzy data trees. f’ is a sub-tree of F, written f’ / F, when
 V’ # V, w’ = w \ V’  V’.
 if i 2 V’ and (j, si) 2 w, then j 2 V’.
 ‘’ and s’ indicate the restriction of ‘ and s to the nodes in V’,
respectively.
 j’ 2 j.
Definition. Let fuzzy data trees f1 = (V1, w1, ‘1, s1, j1, d1) and
f1 = (V2, w2, ‘2, s2, j2, d2) be the sub-trees of F = (V, w, ‘, s, j, d). f1 and
f2 are isomorphic (recorded f1  f2), when
 V1 [ V2 # V, w1 [ w2 # w and s1 [ s2 # s.
 There is a one-to-one mapping, n‘: ‘1 ? ‘2, which makes
n‘ (‘1) = ‘2.
Several fuzzy constructs have been introduced for fuzzy XML data
modeling in Ma and Yan (2007). To accommodate these fuzzy con-
structs, the DTD of the source XML document is correspondingly
modified. First Val element is defined as follows:
<!ELEMENT Val (#PCDATA| original-definition)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
Then Dist element is defined as follows:
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive|conjunctive) ‘‘disjunctive’’>
For a leaf element which only contains text or #PCDATA, say, lea-
fElement, its definition in the DTD is changed from
<!ELEMENT leafElement (#PCDATA)> to
<!ELEMENT leafElement (#PCDATA | Dist)>.
That is, leaf element leafElement may be a crisp one (e.g., sname of
student in Fig. 1), and then could be defined as
<!ELEMENT leafElement (#PCDATA)>.
Also, it is possible that leaf element leafElement may be a fuzzy
one, taking a value represented by a possibility distribution (e.g.,
age of student in Fig. 1). Then it may be defined as
<!ELEMENT leafElement (Dist)>.
Furthermore, we have
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive|conjunctive) ‘‘disjunctive’’>
μES (Ei)/Ei μ RS (Rh)/Rh
(a) (b)
(a) Entity type with membership degree
(c) Single-valued attribute with membership degree
Fig. 1. First Level of Fuzziness in the Fuzzy EER Model.
3617
<!ELEMENT Val (#PCDATA)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
For the non-leaf element, say nonleafElement, first we change
the element definition from < !ELEMENT nonleafElement (origi-
nal-definition)> to
<!ELEMENT nonleafElement (original-definition| Val+| Dist)>
and then add
<!ELEMENT Val (original-definition)>
That is, the non-leaf element nonleafElement may be crisp (e.g.,
student in Fig. 1) and then may be defined as
<!ELEMENT nonleafElement (original-definition)>
When the non-leaf element nonleafElement is a fuzzy one, we dif-
ferentiate two situations: the element takes a value connected
with a possibility degree (e.g., university in Fig. 1), and, second,
the element takes a set of values and each value is connected with
a possibility degree (e.g., employee in Fig. 1). The former element is
defined as follows.
<!ELEMENT nonleafElement (Val+)>
<!ELEMENT Val (original-definition)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
The later element is defined as
<!ELEMENT nonleafElement (Dist)>
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive|conjunctive)
‘‘disjunctive’’>
<!ELEMENT Val (original-definition)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
2.2. Fuzzy EER model
Formally an EER model consists of a set of entity types ES = {E1,
E2, . . ., Ek} and a set of relationship types RS = {R1, R2, . . ., Ru}. For Ei -
2 ES (0 6 i 6 k), AE (Ei) means a set of attributes that entity type Ei
contains, denoted AE (Ei) = {Ai1, Ai2, . . ., Aim}, and IE (Ei) means a
set of entity instances that entity type Ei contains, denoted IE
(Ei) = {ei1, ei2, . . ., ein}. Consequently, Aj (Ei) is used to denote the
attribute Aij of Ei and ep (Ei) is used to denote the entity eip of Ei
(0 6 i 6 k, 1 6 j 6 m and 1 6 p 6 n). Also ep (Aj (Ei)) is used to denote
the value of entity eip on attribute Aij, where eip and Aij are an entity
instance and an attribute of entity type Ei, respectively. In the
context of the given Ei, ep (Aj) may be used instead of ep (Aj (Ei)).
For Rh 2 RS (0 6 h 6 u), AR (Rh) means a set of attributes that rela-
tionship type Rh contains, denoted AR (Rh) = {Ah1, Ah2, . . ., Ahv}, and
IR (Rh) means a set of relationship instances that relationship type
Rh contains, denoted IR (Rh) = {ri1, ri2, . . ., riw}. Consequently, Al
μ AE (Aj)/Aj μ AE (Aj)/Aj
(c) (d)
(b) Relationship type with membership degree
(d) Multivalued attribute with membership degree
3618 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
(Rh) is used to denote the attribute Ahl of Rh and rq (Rh) is used to de-
note the relationship rhq of Rh (0 6 h 6 u, 1 6 l 6 v and 1 6 q 6 w).
Also rq (Al (Rh)) is used to denote the value of relationship rhq on
attribute Ahl, where rhq and Ahl are a relationship instance and an
attribute of relationship type Rh, respectively. In the context of
the given Rh, rq (Rh) may be used instead of rq (Al (Rh)).
To accommodate fuzzy information in the EER model, the EER
data model should be extended by using the fuzzy set and fuzzy lo-
gic (Zadeh, 1978). Generally speaking, in a fuzzy EER model, the ES,
RS, AE (Ei), IE (Ei), AR (Rh) and IR (Rh) may be fuzzy sets, and ep (Aj
(Ei)) and rq (Al (Rh)) may be fuzzy values. In the following we inves-
tigate the details of the fuzzy EER data model.
2.2.1. Fuzzy entity type, fuzzy relationship and fuzzy attribute
Three levels of fuzziness are formally described in the context of
the entity types and attributes. The first level of fuzziness is the
fuzziness by the extent to which the entity type or relationship
type belongs to the data model as well as fuzziness on the content
(in terms of attributes) of the type. The second level of fuzziness is
the fuzziness related to the fuzzy occurrences of entities and rela-
tionships. Even though the structure of an entity type or a relation-
ship type is crisp, it is possible that an instance belongs to the
corresponding type with a membership degree. The third level of
fuzziness is the fuzziness is on attribute values of entities and rela-
tionships. An attribute in an entity or a relationship defines a value
domain. When this domain is a fuzzy subset or a set of fuzzy sub-
set, the fuzziness of an attribute value appears.
Formally, for an entity type, say Ei (1 6 i 6 k), we have lES (Ei)/Ei,
where 0 6 lES (Ei) 6 1 is the degree of Ei belonging to ES; for an
attribute of Ei, say Aj (Ei) (1 6 j 6 m), we have lAE (Aj (Ei))/Aj (Ei),
where 0 6 lAE (Aj (Ei)) 6 1 is the degree of Aj (Ei) belonging to AE
(Ei). Also for an relationship type, say Rh (1 6 h 6 u), we have lRS
(Rh)/Rh, where 0 6 lRS (Rh) 6 1 is the degree of Rh belonging to
RS; for an attribute of Rh, say Al (Rh) (1 6 l 6 v), we have lAR (Al
(Rh))/Al (Rh), where 0 6 lAR (Al (Rh)) 6 1 is the degree of Al (Rh)
belonging to AR (Rh). To model the first level of fuzziness, the name
of attribute, entity type or relationship type should follow a pair of
words ‘‘mem/’’, where 0 < mem 6 1 and it is used to explicitly indi-
cate the membership degree that the attribute belongs to the cor-
responding entity type or relationship degree. The graphical
representations of the first level of fuzziness in the fuzzy EER mod-
el are shown in Fig. 1.
For an entity instance of entity type Ei, say ep (Ei) (1 6 p 6 n),
formally we have lIE (ep (Ei))/ep (Ei), where 0 6 lIE (ep (Ei)) 6 1 is
the degree of ep (Ei) belonging to IE (Ei). Also for a relationship in-
stance of relationship type Rh, say rq (Rh) (1 6 q 6 w), we have lIR
(rq (Rh))/rq (Rh), where 0 6 lIR (rq (Rh)) 6 1 is the degree of rq (Rh)
belonging to IR (Rh). For the second level of fuzziness, an additional
attribute is introduced into the corresponding entity type or rela-
tionship type to represent such a case. This special attribute is de-
noted by l and its attribute domain is [0, 1]. In order to
differentiate the entity type with the second level of fuzziness, a
dashed-outline rectangle is used to denote it. Similarly a dashed-
outline diamond is used to denote the relationship type with the
second level of fuzziness. The graphical representations of the sec-
ond level of fuzziness in the fuzzy EER model are shown in Fig. 2.
For the value of entity on the given attribute, say ep (Aj (Ei))
(1 6 p 6 n), its fuzzy value is formally represented by a possibility
distribution (a fuzzy set), say {p (x1)/x1, p (x2)/x2, . . ., p (xf)/xf}.
Ei
Rh
μ μ
(a) (b)
Fig. 2. The Second Level of Fuzziness in the Fuzzy EER Model.
Here, p (xs) (1 6 s 6 f) denotes the possibility that ep (Aj (Ei)) has va-
lue xs. For the value of relationship on the given attribute, say rq (Al
(Rh)) (1 6 q 6 w), it is a fuzzy value and is represented by a possi-
bility distribution (a fuzzy set), say {p (y1)/y1, p (y2)/y2, . . ., p (yg)/
yg}. Here, p (yt) (1 6 t 6 g) denotes the possibility that rq (Al (Rh))
has value yt. The third level of fuzziness means that attributes
may take fuzzy values. Since an attribute may be single-valued
or multivalued, we have two interpretations to a fuzzy attribute
value: a disjunctive one and a conjunctive one. A dashed-outline
ellipse is used to denote the attribute taking fuzzy value(s). The
graphical representations of the third level of fuzziness in the fuzzy
EER model are shown in Fig. 3.
It is possible that the first level of fuzziness and the third level
of fuzziness appear in an attribute simultaneously. Then we have
disjunctive fuzzy value attribute with membership degree and
conjunctive fuzzy attribute with membership degree. Also it is
possible hat the first level of fuzziness and the second level of fuzz-
iness appear in an entity type and a relationship type simulta-
neously, and then we have entity type with membership degree
as well as fuzzy entities, and relationship type with membership
degree as well as fuzzy relationships. The graphical representations
of the hybrid fuzziness in the fuzzy EER model are shown in Fig. 4.
2.2.2. Fuzzy constraints of relationships
The constraint of a relationship among entity types in the EER
model is called cardinality ratio constraint. Also in the EER model
there is participation constraint, which stipulates the way that enti-
ties participate in the relationship. The concept of participation de-
gree is used to express the minimum number and maximum
number of an entity participating in a relationship, expressed as
(min, max) formally, where max P min P 0 and max P 1. When
min = 0, the way of an entity participates in a relationship is called
partial participation, and is called total participation otherwise. The
cardinality ratio constraint and participation constraint are, some-
times, referred to as the structure constraint.
In the fuzzy ER model, an entity may have the second level of
fuzziness and be associated with a degree that the entity belongs
to the entity type. Also the first level of fuzziness may exist in Ei
and Ej and relationship type Rij, namely, they are the fuzzy entity
types with membership degrees and the fuzzy relationship type
with membership degree. Then The participation constraints and
the cardinality constraints may be fuzzy. Formally, let Rij be a rela-
tionship type of entity types Ei and Ej, denoted Rij (Ei, Ej), and let
degree_Ei/Ei, degree_Ej/Ej, and degree_Rij/Rij. Let b be a threshold.
Then Ei totally b-participates in Rij if
ð8ei Þð9ej ÞðlEi ðei Þ  b ^ lEj ðej Þ  b ^ lRij ðr ij ðei ; ej ÞÞ  b ^ degree Ei
 b ^ degree Ej  b ^ degree Rij bÞ:
And Ei partially b-participates in Rij if
ð9ei Þð9ej ÞðlEi ðei Þ  b ^ lEj ðej Þ  b ^ 0 < lRij ðr ij ðei ; ej ÞÞ < b ^ degree Ei
 b ^ degree Ej  b ^ degree Rij  bÞ:
Here ei and ej are the entities of Ei and Ej with instance membership
degrees lEi (ei) and lEj (ej), respectively, and rij (ei, ej) is a relation-
ship of entities ei and ej. The fuzzy participation means that every
ei of Ei with membership degree being greater than 0 participates
in at least one relationship of Rij with a membership degree being
greater than 0 as well.
(a) Entity type with fuzzy entities
(b) Relationship type with fuzzy relationships
 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627 3619

belonging to the subtype must belong to the supertype. In the fuz-

(a) Disjunctive fuzzy value attribute
(b) Conjunctive fuzzy value attribute
(a) (b)
Fig. 3. Third Level of Fuzziness in the Fuzzy EER Model.
The graphical representations of the fuzzy participation, total
b-participation and partial b-participation in the fuzzy EER model
are shown in Fig. 5.
The cardinality constrain refers to the correspondence between
the numbers of the related entity types and one-to-one correspon-
dence (1:1), one-to-many correspondence (1:n), and many-to-many
correspondence (m:n) can be identified. In the fuzzy ER model, fuzzy
one-to-many correspondence is denoted 1:Ñ, for entity types Ei and
Ej with relationship type Rij, where Ñ is a fuzzy set representing
‘‘how many’’ in the ‘‘many’’ side. That means that for each entity
ei of Ei, there may exist as many as Ñ entities of Ej corresponding
to ei, and for each entity ej of Ej, there is at most one entity of Ei cor-
responding to ej. Furthermore there are two kinds of fuzzy many-to-
many correspondence, denoted m:Ñ and Õ:Ñ, respectively, for entity
types Ei and Ej with relationship type Rij, where Õ and Ñ are fuzzy
sets. Here m:Ñ means that for each entity ej of Ej, there may exist
more than one entity of Ei corresponding to ej, and for each entity
ei of Ei, there may exist as many as Ñ entities of Ej corresponding to
ei. Õ:Ñ means that for each entity ei of Ei, there may exist as many
as Ñ entities of Ej corresponding to ei, and for each entity ej of Ej,
there may exist as many as Õ entities of Ei corresponding to ej.
The graphical representations of the fuzzy cardinality ratios in
the fuzzy EER model are shown in Fig. 6.
2.2.3. Fuzzy specialization and fuzzy generalization
A subtype is produced from supertype by means of inheriting
some attributes of the supertype, overriding some attributes of
the supertype, and defining some new attributes. A subtype is
the specialization of the supertype and any one entity instance
zy EER model, entity types may be fuzzy ones with the second level
of fuzziness as well as the first level of fuzziness. So the subtype/
supertype relationship is fuzzy. In other words, an entity type is
a subclass of another entity type with membership degree of
[0, 1]. The following rules are used to determine the fuzzy sub-
type/supertype relationship.
(a) For any (fuzzy) entity, if the instance membership degree
that it belongs to the subtype is less than or equal to the
instance membership degree that it belongs to the
supertype,
(b) the instance membership degree that it belongs to the sub-
type and the supertype is greater than or equal to a given
threshold, and
(c) the membership degree of the supertype is greater than or
equal to the membership degree of the subtype.
The subtype is then a subtype of the supertype with the mem-
bership degree.
Formally let A be a fuzzy supertype of fuzzy subtypes B1, B2, . . .,
Bn. Without loss of generality, suppose that these entity types have
the instance membership degrees lA, lB1, lB2, . . ., and lBn, and have
the membership degrees degree_A, degree_B1, degree_B2, . . ., and
degree_Bn, respectively. Let b be a given threshold. Then we have
ð8eÞðb 6 maxðlB1 ðeÞ; lB2 ðeÞ; . . . ; lBn ðeÞÞ 6 lA ðeÞÞ ^ b
6 maxðdegree B1 ; degree B2 ; . . . ; degree Bn Þ 6 degree AÞ
That means that B1, B2, . . ., Bn are subtypes of A only if, in addition
that the membership degrees of all entities belonging to A and B1,
B2, . . ., Bn must be greater than or equal to the given threshold
and the membership degree of any entity belonging to A must be
greater than or equal to the membership degree of this entity
belonging to B1, B2, . . ., Bn, the membership degrees of A and B1,
B2, . . ., Bn must be greater than or equal to the same given threshold

μES (Ei)/Ei μ RS (Rh)/Rh μ AE (Aj)/Aj μ AE (Aj)/Aj

(a) (b) (c) (d)

(a) Entity type with membership degree and fuzzy entities (b) Relationship type with membership degree and fuzzy relationships
(c) Disjunctive fuzzy value attribute with membership degree (d) Conjunctive fuzzy value attribute with membership degree

Fig. 4. Hybrid Fuzziness in the Fuzzy EER Model.

Rij β Rij β Rij

Ei Ei Ei

(a) (b) (c)

β β
μ (Ei)/Ei μ (Rij)/Rij μ (Ei)/Ei μ (Rij)/Rij μ (Ei)/Ei μ (Rij)/Rij

(d) (e) (f)

(a) Fuzzy participation of Ei in Rij (b) Total β- participation of Ei in Rij (c) Partial β- participation of Ei in Rij
(d) Fuzzy participation of Ei in Rij with the first level of fuzziness (e) Total β- participation of Ei in Rij with the first level of fuzziness
(f) Partial β- participation of Ei in Rij with the first level of fuzziness

Fig. 5. Fuzzy Participation in the Fuzzy EER Model.

1 Ñ m Ñ Õ Ñ
Ei Rij Ej Ei Rij Ej Ei Rij Ej

Fig. 6. Fuzzy Cardinality Ratios in the Fuzzy EER Model.

3620 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
and the membership degree of A must be greater than or equal to
the membership degree of B1, B2, . . ., Bn.
Furthermore we have the fuzzy total specialization, fuzzy par-
tial specialization, fuzzy disjoint specialization, and fuzzy overlap-
ping specialization as follows.
(a) B1, B2, . . ., Bn are a fuzzy total specialization of A if
ð8eÞð9BÞðe 2 U ^ B 2 fB1 ; B2 ; . . . ; Bn g ^ b 6 lB ðeÞ 6 lA ðeÞ ^ b
6 degree B 6 degree AÞ:
For any of entity belonging to supertype with an instance member-
ship degree that is not less than the given threshold, it must belong
to one of its subtypes with a membership degree that is not less
than the given threshold as well, and the later membership degree
is not greater than the former membership degree. In addition, the
later membership degree is not greater than the former member-
ship degree.
(b) B1, B2, . . ., Bn are a fuzzy partial specialization of A if
ð9eÞð8BÞðe 2 U ^ B 2 fB1 ; B2 ; . . . ; Bn g ^ ððlA ðeÞ  b ^ lB ðeÞ
< bÞ ^ :ðb 6 degree B 6 degree AÞÞÞ:
That is, there exists an entity that belongs to supertype with a
membership degree being greater than or equal to the given thresh-
old but belongs to any of subtypes with a membership degree being
less than the given threshold.
(c) B1, B2, . . ., Bn are disjoint if
ð9=eÞð8BÞð8B0 Þðe 2 U ^ B 2 fB1 ; B2 ; . . . ; Bn g ^ B0
2 fB1 ; B2 ; . . . ; Bn g ^ minðlB ðeÞ; lB0 ðeÞÞ
 b ^ minðdegree B; degree B0 Þ  bÞ:
This means that there exists no any entity that belongs to more than
one subtype each with a membership degree being greater than or
equal to the given threshold.
(d) B1, B2, . . ., Bn are overlapping if
ð9eÞð8BÞð8B0 Þðe 2 U ^ B 2 fB1 ; B2 ; . . . ; Bn g ^ B0
2 fB1 ; B2 ; . . . ; Bn g ^ minðlB ðeÞ; lB0 ðeÞÞ
 b ^ min ðdegree B; degree B0 Þ  bÞ:
That is, there exists an entity that belongs to more than one subtype
each with a membership degree being greater than or equal to the
given threshold.
For fuzzy multiple inheritance of class, let B be a fuzzy subtype,
A1, A2, . . ., An be its multiple fuzzy supertypes, and they all have
instance membership degrees lB, lA1, lA2, . . ., and lAn, and have
membership degrees degree_B, degree_A1, degree_A2, . . ., and
degree_An, respectively. Then we have
ð8eÞð8AÞðe 2 Uj ^ A 2 fA1 ; A2 ; . . . ; An g ^ b 6 lB ðeÞ 6 lA ðeÞ ^ b
6 degree B 6 minðdegree A1 ; degree A2 ; . . . ; degree An ÞÞ:
The graphical representations of the fuzzy specialization in the
fuzzy EER model are shown in Fig. 7.
d o d o
(a) (b) (c) (d)
Fig. 7. Fuzzy Specialization in the Fuzzy EER Model.
Generalization is the inverse process of specialization. Then we
have fuzzy total generalization, fuzzy disjoint generalization, and
fuzzy overlapping generalization. But there is no fuzzy partial gen-
eralization, which is consistent with the situation in the classical
EER model.
2.2.4. Fuzzy category
A category is a subtype of the union of the supertypes with dif-
ferent entity types. Symbolically, a category B and the supertypes
A1, A2, . . ., An satisfy the relationship B # A1 [ A2 [ . . . [ An. The dif-
ference between a category and a subtype (say B’) with more than
one supertype (say A’1, A’2, . . ., A’m) is that B’ # A’1 \ A’2 . . . A’m.
Compared with the subtype with multiple supertypes, the category
inherits some attributes from one of the supertypes.
Entity types may be fuzzy ones with the second level of fuzzi-
ness as well as the first level of fuzziness. As a result, an entity type
is a category of several entity types with membership degree of
[0, 1]. Formally let B be a fuzzy category of fuzzy entity types A1,
A2, . . ., and An. Without loss of generality, first it is assumed that
the instance membership degrees of these entity types are lB,
lA1, lA2, . . ., and lAn, respectively (the second level of fuzziness);
second it is assumed that these fuzzy entity types have member-
ship degrees degree_B/B, degree_A1/A1, degree_A2/A2, . . ., degree_An/
An, respectively (the first level of fuzziness). Let b be a given thresh-
old. Then B is a category of A1, A2, . . ., and An if
ð8eÞð9AÞðe 2 B ^ A 2 fA1 ; A2 ; . . . ; An g ^ b 6 lB ðeÞ 6 lA ðeÞ ^ b
6 degree B 6 minðdegree A1 ; degree A2 ; . . . ; degree An ÞÞ:
That means that a fuzzy entity type B is the category of a group fuz-
zy entity types A1, A2, . . ., and An if
(a) for any (fuzzy) entity in B, say e, there must exist an entity
type, say Ai (1 6 i 6 n) so that the membership degree that
e belongs to B is less than or equal to the membership degree
that e belongs to entity type Ai,
(b) the instance membership degree that e belongs to B is
greater than or equal to the given threshold, and
(c) for any entity type, say Ai (1 6 i 6 n), its membership degree
degree_Ai is greater than or equal to the degree_B/B, which is
greater than or equal to the given threshold.
Then the membership degree that B is a category of A1, A2, . . .,
and An should be lB (e). The graphical representation of the fuzzy
category in the fuzzy EER model is shown in Fig. 8(a).
∨ ×
(a) Fuzzy category
(b) Fuzzy aggregation
(a) (b)
Fig. 8. Fuzzy Category and Aggregation in the Fuzzy EER Model.
(a) Fuzzy total & disjoint specialization
∧ (b) Fuzzy total & overlapping specialization
11 (c) Fuzzy partial & disjoint specialization
(d) Fuzzy partial & overlapping specialization
(e) Fuzzy subclass with multiple fuzzy superclasses
(e)
 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
2.2.5. Fuzzy aggregation
An aggregation captures a whole-part relationship between an
aggregate and a constituent part and these constituent parts can
exist independently. An entity type aggregated from fuzzy constit-
uent parts must be fuzzy and the aggregation is fuzzy. Then an en-
tity type is an aggregation of constituent parts with membership
degree of [0, 1].
Formally let A be a fuzzy aggregation of fuzzy entity types
B1, B2, . . ., and Bn. Here the instance membership degrees of these
entity types are lA, lB1, lB2, . . ., and lBn, respectively (the second
level of fuzziness). Also these fuzzy entity types have membership
degrees degree_A/A, degree_B1/B1, degree_B2/B2, . . ., degree_Bn/Bn,
respectively (the first level of fuzziness). Let b be a given threshold.
Then A is an aggregate of B1, B2, . . ., and Bn if
ð8eÞð9e1 Þð9e2 Þ... ð9en Þðe 2 A ^ e1 2 B1 ^ e2 2 B2 ^ ... ^ en 2 Bn ^ b 6 lA ðeÞ
6 minðlB1 ðe1Þ; lB2 ðe2Þ;...; lBn ðenÞÞ ^ b 6 degree A
6 minðdegree B1 ;degree B2 ;...;degree Bn ÞÞ:
That means that a fuzzy entity type A is the aggregate of a group
fuzzy entity types B1, B2, . . ., and Bn if
(a) for any (fuzzy) entity in A, say e, there exists a set of entities,
say e1, e2, . . . and en, and the membership degree that e
belongs to entity type A is less than or equal to the member-
ship degree that ei belongs to entity type Bi (1 6 i 6 n),
(b) the instance membership degree that e belongs to A is
greater than or equal to the given threshold, and
(c) for any entity type, say Bi (1 6 i 6 n), its membership degree
degree_Bi is greater than or equal to the degree_A/A, which is
greater than or equal to the given threshold.
The membership degree that A is an aggregation of B1, B2, . . .,
and Bn should be minlBi(ei)Pb (lBi (ei)) (1 6 i 6 n). The graphical
representation of the fuzzy aggregation in the fuzzy EER model is
shown in Fig. 8(b).
3. Mapping fuzzy EER model to fuzzy XML DTD
Using the fuzzy EER model and fuzzy XML model, fuzzy informa-
tion can be modeled at two different levels. This section investigates
the formal mapping of the fuzzy EER model into the fuzzy XML DTD.
3.1. Transformation of entity type
An entity type of EER model generally corresponds to an XML
element type declaration, and the attributes of the entity type in
the EER data model correspond to the element content description
in the XML DTD. Note that, like other conceptual data models, attri-
bute names are mandatory in the EER model whereas the attribute
types are optional. But in the XML, element content only consists of
type names. So it is assumed that attribute names imply their attri-
bute type names (Conrad et al., 2000). When there is no entity type
representing a suitable declaration for an attribute type, the attri-
bute type is assumed to be an element whose content type is
#PCDATA. In addition, multiplicity specifications of attributes are
mapped into cardinality specifications with specifiers ?, ⁄, and +,
which are used for element content construction.
To transform from the fuzzy EER data model into the XML DTD,
we first suppose that the entity types in the fuzzy EER model are
neither subclasses nor superclasses. Then, we can distinguish four
basic kinds of entity types in the fuzzy EER model as follows.
(a) entity types without any fuzziness at the three levels,
(b) entity types with the fuzziness only at the third level,
3621
(c) entity types with the fuzziness at the second level, and
(d) entity types with the fuzziness at the first level.
For the entity types in case (a), they can be transformed following
the approach developed in Conrad et al. (2000). The transformation
of the entity types with the third and second levels of fuzziness is
of particular concern. Formally suppose that we have an entity
type named ‘‘EntityName’’, which contains an attribute named
‘‘AttributeName’’. Here ‘‘EntityName’’ is an entity type which may
contain some entity instances belonging to EntityName with mem-
bership degrees and ‘‘AttributeName’’ is an attribute which may
take fuzzy values. For such an entity type, when its entity type
name is transformed into the name of the element type, the attri-
butes cannot be transformed into element content description
directly. First, we should use
<!ELEMENT EntityName (Dist)>
<!ATTLIST EntityName FID IDREF #REQUIRED>
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive)>
<!ELEMENT Val (AttributeName?)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
rather than directly use
<!ELEMENT EntityName (AttributeName?)>
<!ATTLIST EntityName FID IDREF #REQUIRED>.
Then we should use
<!ELEMENT AttributeName (Dist)>
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive)>
<!ELEMENT Val (#PCDATA)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
to replace
<!ELEMENT AttributeName (#PCDATA)>.
With the fuzzy entity types in (b) and (c) above, a combined kind of
fuzzy entity types (denoted (e)) can be constructed. This kind of fuz-
zy entity types contains fuzziness both at the third and second lev-
els, and their transformation to the fuzzy XML DTD is actually a
simple combination of the above transformations in cases (b) and
(c). For a weak entity type which depends on its owner entity type
in the fuzzy EER model, it can be regarded as a relation-valued attri-
bute and then transformed into the fuzzy DTD. But the entity types
with the first level of fuzziness cannot be mapped into the created
XML DTD because the fuzzy XML DTD model does not support fuzzy
information modeling at a metadata level.
An aggregation represents a whole-part relationship between
an aggregate and a constituent part, in which the constituent parts
can be treated as the special attributes of the aggregate. So the
aggregations can be transformed using the approach to the trans-
formation of entity types.
3.2. Transformation of relationship type
The relationships in the EER model describe connections among
entity types. Basically, for two entity types with of a relationship,
their names become the names of two corresponding element
types, and the attributes of the entity types are transformed into
the corresponding element content descriptions. Then each trans-
formed element must be augmented by a #REQUIRED IDREF attri-
bute, which is an artificial one and comes from another entity type
involved in the relationship.
3622 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627

In the fuzzy EER model, we can distinguish four kinds of rela- <!ELEMENT key-1 (#PCDATA)>
tionship types as follows: <!ELEMENT key-2 (#PCDATA)>
<!ELEMENT attribute (Dist)>
(a) relationship types without any fuzziness at the three levels, <!ELEMENT Dist (Val+)>
(b) relationship types with the fuzziness only at the third level, <!ATTLIST Dist type (disjunctive)>
(c) relationship types with the fuzziness at the second level, and <!ELEMENT Val (#PCDATA)>
(d) relationship types with the fuzziness at the first level. <!ATTLIST Val Poss CDATA ‘‘1.0’’>

Here we only focus on the fuzzy relationship types in cases (b) and
(c). At this point, a fuzzy relationship type is regarded as a fuzzy
entity type, and can be transformed into a virtual element type
in the fuzzy XML DTD. Then the relationship type and entity types
connected with the relationship type are transformed according to
the transformation of fuzzy entity types developed above.
Formally let we have two entity types ‘‘Entity-1’’ and ‘‘Entity-2’’,
which contains primary keys ‘‘key-1’’ and ‘‘key-2’’, respectively.
Also assume that there is a relationship type named ‘‘Relationship-
Type’’ between these two entity types, which has the second level
of fuzziness and contains a primary key ‘‘key-R’’ as well as an attri-
bute ‘‘attribute’’ with the third level of fuzziness. Then ‘‘Entity-1’’
and ‘‘Entity-2’’ can be directly transformed into two separate ele-
ment types following the transformation of fuzzy entity types
developed in Section 3.1. Also ‘‘RelationshipType’’ can be trans-
formed into an element type following the transformation of fuzzy
entity types developed in Section 3.1. But this created element type
must contain ‘‘key-1’’ and ‘‘key-2’’ as its element content descrip-
tion in addition to ‘‘key-R’’ and ‘‘attribute’’, and finally we have
<!ELEMENT RelationshipType (Dist)>
<!ATTLIST RelationshipType yid ID #REQUIRED>
<!ELEMENT Dist (Val+)>
<!ATTLIST Dist type (disjunctive)>
<!ELEMENT Val (key-R, key-1, key-2, attribute)>
<!ATTLIST Val Poss CDATA ‘‘1.0’’>
<!ELEMENT key-R (#PCDATA)>
3.3. Transformation of generalization
The generalization/specialization in the EER model defines a
subtype/supertype relationship between entity types, in which
the entity type called supertype is a more general description of
a set of other entity types called subtypes. Following the same
transformation of entity types, the supertype and all subtypes
are transformed into the element types in the XML DTD, respec-
tively. Here the element type mapped from the supertype is called
a superelement and the element type mapped from a subtype is
called a subelement (Conrad et al., 2000). Note that a superelement
must receive an additional ID attribute stated #REQUIRED, and
each subelement must be augmented by a #REQUIRED IDREF attri-
bute in addition to the transformations that the entity type names
become the names of the element types and the attributes are
transformed into element content description.
Now let us focus on the fuzzy generalization in the fuzzy EER
model. For a fuzzy supertype/subtype in the fuzzy EER model, as-
sume that we have a supertype named ‘‘SuperType’’ and two sub-
types named ‘‘SubType-1’’ and ‘‘SubType-2’’, respectively, the
generalization has fuzziness at the type/instance level as well as
at the attribute value level. Following the transformation of fuzzy
entity types developed in Section 3.1 (i.e., transforming the entity
types in the cases of (b) and (c) above), these three entity types are

Gauss d Ele_Num Num_gauss Num_node Num_DOF Num total_point

VectorNum

composed Element Num_ele_node

Ele_Type
Lnods
VectorNum
Dim μ Dim_xyz

Vector Dvolu d Num_element

Num_nodal_point
form Array composed

Elem_DOF Num_ele_gauss
composed composed Num_Element General_Data

μ Num_bc_point
Total_DOF
form Rdim Element_SEQ Els_Num ID μ

Cdim
Dim_xyz Title Node_DOF Num_gauss
Matrix μ Status Num_Node
MatrixNum

Shape Seq NodeNum Node NumDOF

composed

0.5/Seq node_num NumDim Status xyz disp

μ

Fig. 9. Fuzzy EER Model for Finite Element Analysis.

 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627 3623

Fig. 10. Fuzzy XML DTD Element Types Trnasfromed from the Independent Entity Types in Fig. 9.
3624 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
transformed first. Then the created superelement should be associ-
ated with ID #REQUIRED, and the created each subelement should
be associated with IDREF #REQUIRED. Formally we have
<!ELEMENT SubType -1 (......)>
<!ATTLIST SubType -1 sid IDREF REQUIRED>
......
<!ELEMENT SubType -2 (......)>
<!ATTLIST SubType -2 sid IDREF REQUIRED >
......
<!ELEMENT SuperType (SubType -1?, SubType -2?)>
<!ATTLIST SuperType eid ID #REQUIRED>
......
Categorization can be regarded as a kind of selective inheritance.
The entity types in the categorization, fuzzy or not, can be mapped
into the element types in the fuzzy XML DTD, respectively, follow-
ing the transformation presented above.
4. A case study
The EER model, being an important kind of conceptual data
model, has been traditionally applied in many application domains
such as engineering applications (Eastman and Fereshetian, 1994).
In (Yu and Adeli, 1993), for example, the EER model is used to mod-
el a myriad of data types generated in finite element analysis of
complex engineering systems. Also nowadays XML has been
widely applied in Web-based engineering applications for product
data exchange and integration (Xu and Liu, 2003). A XML-based
product data markup language is proposed in Burkett et al.
(1999). Engineering applications are typically knowledge-intensive
and engineering data is inherently imprecise and uncertain.
Viewed from conceptual design phase of products, for example,
the early phases of product design, the design space is relatively
wide open and some constraints have not been identified. Product
preliminary design is essentially a process of reducing the
Fig. 11. Fuzzy XML DTD Element Types Trnasfromed from the Supertypes in Fig. 9.
imprecision and uncertainty in the description of the conceptual
design (Wood and Antonsoon, 1992). Actually fuzzy set theory
can be used in many engineering applications, where either crisp
information is not available or information flexible processing is
necessary. In order to model fuzzy engineering data, in Yan and
Ma (2013), the IDEF1X model, a conceptual model designed espe-
cially for engineering data modeling, is extended to the fuzzy
IDEF1X model for conceptually designing the fuzzy object-oriented
databases.
In this section, we give an example from engineering applica-
tions to illustrate the transformation of the fuzzy EER model into
the fuzzy XML DTD. We use the fuzzy EER model shown in Fig. 9
rather than the fuzzy IDEF1X model in Yan and Ma (2013) to model
fuzzy engineering information for finite element analysis of engi-
neering problems.
Fig. 9 contains eight entity typea. Among these entity types, en-
tity types Vector, Matrix, General_Data, Element_SEQ and Node are
the entity types with the second level of fuzziness (entity type
Node, for example, may have some entity instances which fuzzily
belong to Node), and entity types Vector, Matrix, General_Data, Ele-
ment_SEQ and Node are the entity types with the third level of fuzz-
iness (entity type Node, for example, may have fuzzy values on
attribute Status). In addition, entity type Shape may or may not
contain attribute seq, and it is an entity type with the first level
of fuzziness on attributes. In addition, there are two relationships,
all named composed, between Element and Vector, and Element and
Matrix; there are three relationships with the second level of fuzz-
iness, all named composed, between Matrix and Vector, Shape and
Matrix, and Element and Element_SEQ. Among these eight entity
types, entity types Gauss and Element have a supertype/subtype
relationship, and entity types General_Data and Element have a fuz-
zy supertype/subtype relationship.
For the transformation of the fuzzy EER model in Fig. 9 to the
fuzzy XML DTD data model, first we need to identify independent
(fuzzy) entity types. An independent (fuzzy) entity type means
such an entity type which is not involved in any supertype/subtype
relationships. An independent (fuzzy) entity type can be trans-
formed into an element type in the (fuzzy) XML DTD directly. In
 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627 3625

Fig. 9, entity types Node, Element_SEQ, Sharp, Matrix and Vector are
identified to be independent (fuzzy) entity types and they can be
directly transformed into element types NODE, ELEMENT_SEQ,
SHARP, MATRIX and VECTOR, respectively. Note that, among these
independent entity types, Element_SEQ is a fuzzy entity type with
the second level of fuzziness, Node, Matrix and Vector are fuzzy en-
tity types with the third level of fuzziness, and these independent
fuzzy entity types can be transformed only following the discus-
sion given in Section 3.1. Finally the XML DTD element types,
which are transformed from the independent (fuzzy) entity types
in Fig. 9, are shown in Fig. 10.
After the independent (fuzzy) entity types are transformed, now
we can focus on the (fuzzy) entity types involved in subtype/super-
type relationships. In Fig. 9, entity type Element is a subtype of
supertypes General_Data and Gauss, respectively, and these three
entity types can be transformed into element types ELEMENT,
GAUSS and GENERAL_DATA in the (fuzzy) XML DTD following the
discussion given in Section 3.3. Entity types Gauss and Gen-
eral_Data are transformed into XML DTD element types GAUSS
and GENERAL_DATA shown in Fig. 11, and entity type Element is
transformed into XML DTD element type ELEMENT shown in
Fig. 12.
Finally we can transform the (fuzzy) relationships. In Fig. 9,
there are two relationships named composed between Element
and Vector, and Element and Matrix; there are three relationships
with the second level of fuzziness named composed between Matrix
and Vector, Shape and Matrix, and Element and Element_SEQ. These
five relationships can be respectively transformed into element
types COMPOSED-1, COMPOSED-2, COMPOSED-3, COMPOSED-4,
and COMPOSED-5 in the (fuzzy) XML DTD following the discussion

Fig. 12. Fuzzy XML DTD Element Type Trnasfromed from the Subtype in in Fig. 9.

Fig. 13. Fuzzy XML DTD Element Type Trnasfromed from the Relationships in in Fig. 9.
3626 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
given in Section 3.2. These transformed element types are shown
in Fig. 13.
In the transformation discussed above, the fuzziness in the fuz-
zy EER model is preserved in the fuzzy XML DTD data model except
for the fuzziness in entity type Sharp, where attribute ‘‘Seq’’ bongs
to Sharp with membership degree 0.5. After the transformations
above, the fuzzy EER data model in Fig. 9 is finally transformed into
the fuzzy XML DTD data model.
5. Conclusion
Imprecise and uncertain data exist in many the real-world
applications and imprecise and uncertain data modeling has been
extensively investigated mainly in the context of classical database
models (e.g., relational database model). Recently, with the prompt
development and wide applications of the Web, XML has been the
de facto standard of data representation and exchange on the Web
and imprecise and uncertain XML data management has received
more attention. This paper investigates the formal mapping from
the fuzzy conceptual data model to the fuzzy XML model. The fuz-
zy conceptual data model used in the paper is the fuzzy EER model,
a fuzzy extension to the classical EER model which is widely ap-
plied in information modeling. Both the fuzzy EER model and the
fuzzy XML model are formally described in the paper. On the basis,
the formal approach to translating the fuzzy EER model into the
fuzzy XML model is developed. The proposed approach can serve
as the conceptual design of fuzzy XML model and fuzzy data can
be modeled in two kinds of different data models. The formal
translation from the fuzzy XML model to the fuzzy EER model is
developed in our previous work (Yan and Ma, 2012).
It should be noted that the fuzzy XML model discussed in this
paper only focuses on the fuzzy XML DTD. Compared with XML
DTD which lacks enough expressive power to properly describe
highly structured data, XML Schema provides a much richer set
of structures, types and constraints for describing data. So in the
near future, we will concentrate on the fuzzy XML Schema model,
investigating the formal translation from the fuzzy EER model to
the fuzzy XML Schema model as well as the reengineering of the
fuzzy XML Schema model into the fuzzy EER model.
Acknowledgements
The authors wish to thank the anonymous referees for their
valuable comments and suggestions, which improved the technical
content and the presentation of the paper. The work is supported
by the National Natural Science Foundation of China (61370075,
60873010 and 61073139) and the Program for New Century Excel-
lent Talents in University (NCET-05-0288).
References
Abiteboul, S., Segoufin, L. & Vianu, V. (2001). Representing and Querying XML with
Incomplete Information. In Proceedings of the 12th ACM SIGACT-SIGMOD-SIGART
symposium on principles of database systems (pp. 150–161).
Aygun, R. S., & Yazici, A. (2004). Modeling and management of fuzzy information in
multimedia database applications. Multimedia Tools Applications, 24(1), 29–56.
Bernauer, M., Kappel, G. & Kramler, G. (2004). Representing XML Schema in UML-A
Comparison of Approaches. In Proceedings of the 4th international conference on
web engineering (Vol. 3140, pp. 440–444). Lecture notes in computer science.
Berlin Heidelberg: Springer-Verlag. .
Bertino, E., & Catania, B. (2001). Integrating XML and databases. IEEE Internet
Computing, 84–88. July–August.
Bordogna, G., Pasi, G., & Lucarella, D. (1999). A fuzzy object-oriented data model for
managing vague and uncertain information. International Journal of Intelligent
Systems, 14, 623–651.
Buckles, B. P., & Petry, F. E. (1982). A fuzzy representation of data for relational
database. Fuzzy Sets and Systems, 7(3), 213–226.
Burkett, W. (1999). PDML: A new paradigm for product data exchange and
integration. White paper. http://www.pdit.com/pdml/whitepap.pdf.
Chamorro-Martínez, J., Medina, J. M., Barranco, C. D., Galán-Perales, E., & Soto-
Hidalgo, J. M. (2007). Retrieving images in fuzzy object-relational databases
using dominant color descriptors. Fuzzy Sets and Systems, 158(3), 312–324.
Chen, G. Q. & Kerre, E. E. (1998). Extending ER/EER Concepts towards Fuzzy
Conceptual Data Modeling. In Proceedings of the 1998 IEEE international
conference on fuzzy systems (Vol. 2, pp. 1320–1325).
Conrad, R., Scheffner, D. & Freytag, J. C. (2000). XML Conceptual Modeling Using
UML. In Proceedings of the 19th International Conference on Conceptual Modeling
(Vol. 1920, pp. 558–571). Lecture notes in computer science. Berlin Heidelberg:
Springer-Verlag.
Cuevasa, L., Marínb, N., Ponsb, O., & Vilab, M. A. (2008). Pg4DB: A fuzzy object-
relational system. Fuzzy Sets and Systems, 159(12), 1500–1514.
Dubois, D., Prade, H., & Rossazza, J. P. (1991). Vagueness, typicality, and uncertainty
in class hierarchies. International Journal of Intelligent Systems, 6, 167–183.
Eastman, C. M., & Fereshetian, N. (1994). Information models for use in product
design: A comparison. Computer-Aided Design, 26(7), 551–572.
Elmasri, R., Li, Q., Fu, J., Wu, Y.-C., Hojabri, B., & Ande, S. (2005). Conceptual modeling
for customized xml schemas. Data & Knowledge Engineering, 54(1), 57–76.
Galindo, J., Urrutia, A., Carrasco, R. A., & Piattini, M. (2004). Relaxing constraints in
enhanced entity-relationship models using fuzzy quantifiers. IEEE Transactions
on Fuzzy Systems, 12(6), 780–796.
Gaurav, A., Alhajj, R. (2006). Incorporating fuzziness in XML and mapping fuzzy
relational data into fuzzy XML. In Proceedings of the 2006 ACM symposium on
applied computing (pp. 456–460).
George, R., Srikanth, R., Petry, F. E., & Buckles, B. P. (1996). Uncertainty management
issues in the object-oriented data model. IEEE Transactions on Fuzzy Systems,
4(2), 179–192.
Gyseghem, N. V., & Caluwe, R. D. (1998). Imprecision and uncertainty in UFO
database model. Journal of the American Society for Information Science, 49(3),
236–252.
Hung, E., Getoor, L., & Subrahmanian, V. S. (2003). PXML: A probabilistic
semistructured data model and algebra. In Proceedings of the 19th the
international conference on data engineering (pp. 467–478).
Ma, Z. M. (2005). A conceptual design methodology for fuzzy relational databases.
Journal of Database Management, 16(2), 66–83.
Ma, Z. M., Liu, J., & Yan, L. (2010). Fuzzy data modeling and algebraic operations in
XML. International Journal of Intelligent Systems, 25(9), 925–947.
Ma, Z. M., & Shen, D. (2006). Modeling fuzzy information in the IF2O and object-
oriented data models. Journal of Intelligent & Fuzzy Systems, 17(6), 597–612.
Ma, Z. M., & Yan, L. (2007). Fuzzy XML data modeling with the UML and relational
data models. Data & Knowledge Engineering, 63, 972–996.
Ma, Z. M., & Yan, L. (2008). A literature overview of fuzzy database models. Journal of
Information Science and Engineering, 24(1), 189–202.
Ma, Z. M., & Yan, L. (2010). A literature overview of fuzzy conceptual data modeling.
Journal of Information Science and Engineering, 26(2), 427–441.
Ma, Z. M., Yan, L., & Zhang, F. (2012). Modeling fuzzy information in UML class
diagrams and object-oriented database models. Fuzzy Sets and Systems, 186(1),
26–46.
Ma, Z. M., Zhang, Fu, & Yan, Li (2011). Fuzzy information modeling in UML class
diagram and relational database models. Applied Soft Computing, 11(6),
4236–4245.
Ma, Z. M., Zhang, W. J., & Ma, W. Y. (2004). Extending object-oriented databases for
fuzzy information modeling. Information Systems, 29(5), 421–435.
Ma, Z. M., Zhang, W. J., Ma, W. Y., & Chen, G. Q. (2001). Conceptual design of fuzzy
object-oriented databases using extended entity-relationship model.
International Journal of Intelligent Systems, 16, 697–711.
Ma, Z., & Yan, L. (2010). Soft computing in XML data management. Springer-Verlag.
Mani, M., Lee, D. W. & Muntz, R. R. (2001). Semantic data modeling using XML
schemas. In Proceedings of the 20th international conference on conceptual
modeling (Vol. 2224, pp. 149–163). Lecture notes in computer science. Berlin
Heidelberg: Springer-Verlag.
Nierrman, A. & Jagadish, H. V. (2002). ProTDB: Probabilistic data in XML. In Proceedings
of the 28th international conference on very large data bases (pp. 646–657).
Oliboni, B. & Pozzani, G. (2008). Representing fuzzy information by using XML
schema. In Proceedings of the 2008 international conference on database and
expert systems application (pp. 683–687).
Petrovic, D., Roy, R., & Petrovic, R. (1999). Supply chain modeling using fuzzy sets.
International Journal of Production Economics, 59, 443–453.
Prade, H., & Testemale, C. (1984). Generalizing database relational algebra for the
treatment of incomplete or uncertain information and vague queries.
Information Sciences, 34, 115–143.
Psaila, G. (2000). ERX: A data model for collections of XML documents. In
Proceedings of the 2000 ACM symposium on applied computing (Vol. 2, pp. 898–
903).
Raju, K. V. S. V. N., & Majumdar, K. (1988). Fuzzy functional dependencies and
lossless join decomposition of fuzzy relational database systems. ACM
Transactions on Database Systems, 13(2), 129–166.
Senellart, P. & Abiteboul, S. (2007). On the complexity of managing probabilistic
XML data. In Proceedings of the 26th ACM SIGMOD-SIGACT-SIGART symposium on
principles of database systems (pp. 283–292).
Umano, M., & Fukami, S. (1994). Fuzzy relational algebra for possibility-
distribution-fuzzy-relational model of fuzzy data. Journal of Intelligent
Information Systems, 3, 7–27.
Van Keulen, M., De Keijzer, A. & Alink, W. (2005). A probabilistic XML approach to
data integration. In Proceedings of the 2005 international conference on data
engineering (pp. 459–470).
 L. Yan, Z.M. Ma / Expert Systems with Applications 41 (2014) 3615–3627
Wood, K. L., & Antonsoon, E. K. (1992). Modeling imprecision and uncertainty in
preliminary engineering design. Mechanism and Machine Theory, 25(3),
305–324.
Xiao, R. G., Dillon, T. S., Chang, E. & Feng, L. (2001). Modeling and transformation of
object-oriented conceptual models into XML schema. In Proceedings of the12th
international conference on database and expert systems applications (Vol. 2113,
pp. 795–804). Lecture notes in computer science. Berlin Heidelberg: Springer-
Verlag.
Xu, X., & Liu, D. T. (2003). A web-enabled PDM system in collaborative design
environment. Robotics and Computer-Integrated Manufacturing, 19(4), 315–328.
Yager, R. R. (2000). Targeted e-commerce marketing using fuzzy intelligent agents.
IEEE Intelligent Systems, 15(6), 42–45.
Yager, R. R., & Pasi, G. (2001). Product category description for web-shopping in e-
commerce. International Journal of Intelligent Systems, 16, 1009–1021.
Yan, L., & Ma, Z. M. (2012). Incorporating fuzzy information into the formal mapping
from web data model to extended entity-relationship model. Integrated
Computer-Aided Engineering, 19(4), 313–330.
3627
Yan, L., & Ma, Z. M. (2013). Conceptual design of object-oriented databases for fuzzy
engineering information modeling. Integrated Computer-Aided Engineering,
20(2), 183–197.
Yan, L., Ma, Z. M. & Liu, J. (2009). Fuzzy data modeling based on XML schema. In
Proceedings of the 2009 ACM international symposium on applied computing (pp.
1563–1567).
Yazici, A., Buckles, B. P., & Petry, F. E. (1999). Handling complex and uncertain
information in the ExIFO and NF2 data models. IEEE Transactions on Fuzzy
Systems, 7(6), 659–676.
Yu, G., & Adeli, H. (1993). Object-oriented finite element analysis using EER model.
Journal of Structural Engineering, 119(9), 2763–2781.
Zadeh, L. A. (1965). Fuzzy Sets. Information and Control, 8(3), 338–353.
Zadeh, L. A. (1978). Fuzzy sets as a basis for a theory of possibility. Fuzzy Sets and
Systems, 1(1), 3–28.
Zvieli, A. & Chen, P. P. (1986). Entity-relationship modeling and fuzzy databases. In
Proceedings of the 1986 IEEE international conference on data engineering (pp.
320–327).