Automation in Construction 57 (2015) 202–212

Contents lists available at ScienceDirect

Automation in Construction

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

A fuzzy extension of the semantic Building Information Model

J. Gómez-Romero a,b,⁎, F. Bobillo c, M. Ros a,b, M. Molina-Solana a,b, M.D. Ruiz a,b, M.J. Martín-Bautista a,b
a
Department of Computer Science and Artificial Intelligence, University of Granada, Spain
b
Research Centre for Information and Communications Technologies (CITIC-UGR), Spain
c
Department of Computer Science and Systems Engineering, University of Zaragoza, Spain

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

Article history: The Building Information Model (BIM) has become a key tool to achieve communication during the whole build-
Received 1 November 2014 ing life-cycle. Open standards, such as the Industry Foundation Classes (IFC), have contributed to expand its
Received in revised form 3 April 2015 adoption, but they have limited capabilities for cross-domain information integration and query. To address
Accepted 13 April 2015
these challenges, the Linked Building Data initiative promotes the use of ontologies and Semantic Web technol-
Available online 14 May 2015
ogies in order to create more formal and interoperable BIMs. In this paper, we present a fuzzy logic-based
Keywords:
extension of such semantic BIMs that provides support for imprecise knowledge representation and retrieval.
Building Information Model We propose an expressive fuzzy ontology language, and describe how to use a fuzzy reasoning engine in a BIM
Ontologies context with selected examples. The resulting fuzzy semantic BIM enables new functionalities in the project
Fuzzy ontologies design and analysis stages—namely, soft integration of cross-domain knowledge, flexible BIM query, and
imprecise parametric modeling.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction semantics of BIM data models, but also more flexible and with enhanced
query capabilities in order to express different perspectives on building
The use of Building Information Models (BIMs) in the architecture, information, to facilitate information retrieval, and to integrate the BIM
engineering and construction industry has evolved from the early with other information resources. Given its open and neutral character,
three-dimensional blueprints of the building geometry developed in the Industry Foundation Classes (IFC) standard, proposed by the
the 70s to the complex representations of volumes, materials, and buildingSMART alliance [3], has been typically used as the basis
equipment that are nowadays more and more frequent. BIMs have for these extended representations. The IFC specification defines an
proved very effective to increase building quality while reducing design object-based data model written in the EXPRESS data definition
and construction costs by enabling better interoperation between language, and an accompanying text-based file interchange format
stakeholders [1]. based on STEP. It allows creating readable models and data validation
According to the US National Building Information Model Standard rules, but it lacks a mathematical characterization of the semantics
Project Committee [2], “a Building Information Model (BIM) is a digital of its representation primitives. Consequently, querying the model is es-
representation of physical and functional characteristics of a facility.” A sentially an informal procedure supported by ad hoc implementations.
remarkable feature of this definition is that it highlights the relevance of Not surprisingly, Semantic Web technologies have been proposed to
the BIM as a “shared knowledge resource for information about a facility” address the challenges of the next generation BIMs [4], since they offer a
that provides support for decision-making “during its life-cycle”, thus complete framework for the management of the knowledge published
expanding its utilization beyond the design stage. Nevertheless, full- in the Web, arguably the most heterogeneous information environment
fledged BIM applications covering the whole building life-cycle are still of our days. The envisioned Linked Building Data cloud, based on the
scarce, because it implies interfacing with heterogeneous users who Semantic Web technology stack [5], increases interoperability during
have different background, objectives, and priorities. the building life-cycle by connecting distributed pieces of BIM data
For this reason, in the last years there is an increasing interest in the [6,7] and cross-domain data [8]. At the core of the Linked Building
development of knowledge representations able to capture the Data cloud, ontologies encoded in OWL 2 (Ontology Web Language)
[9] are used to define a formal conceptual schema for BIM constituents,
and the RDF (Resource Description Framework) language [10] is used to
⁎ Corresponding author at: Department of Computer Science and Artificial Intelligence, encode BIM instances. We call semantic BIMs to these BIMs represented
University of Granada, Spain.
E-mail addresses: [email protected] (J. Gómez-Romero), [email protected]
in OWL/RDF. The semantic BIM leverages classical BIM query capabili-
(F. Bobillo), [email protected] (M. Ros), [email protected] (M. Molina-Solana), ties by enabling automatic reasoning to retrieve information and to
[email protected] (M.D. Ruiz), [email protected] (M.J. Martín-Bautista). infer implicit knowledge.

http://dx.doi.org/10.1016/j.autcon.2015.04.007
0926-5805/© 2015 Elsevier B.V. All rights reserved.
 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212 203

The theoretical underpinnings of Semantic Web ontologies are 2. Materials, methods and tools
strongly based on Description Logics (DLs), a subset of first order logic
especially suitable for representing structured knowledge [11]. However, 2.1. Ontologies and the BIM
DLs cannot directly manage imprecise knowledge, which is inherent to
several real-world problems [12]. This is the case of the semantic BIM, Ontologies are typically used for representing knowledge in
in which we may like for instance to represent that a building element scenarios that require interoperation between heterogeneous agents.
is quite big, two elements are quite similar, there are several elements As mentioned, this is the case of the BIM, where several individuals
inside a space, and so forth. It would be also convenient to allow querying with different expertise are usually involved. Essentially, an ontology
the system in these same terms; for example, to retrieve all the elements is developed from the following primitive elements: (i) classes
with size around a dimension value, or those that have been built with (or concepts), which determine sets that classify domain objects;
similar materials. (ii) instances (or individuals), which are concrete occurrences of
Fuzzy logic and fuzzy set theory are appropriate formalisms to han- concepts; (iii) properties (also named relations or roles), which represent
dle imprecise knowledge. Hence, several proposals of fuzzy Description binary connections between individuals, or individuals and typed values
Logics to support fuzzy ontologies have emerged [13]. Generally speak- (integers, strings, etc.); and (iv) axioms, which establish restrictions
ing, in fuzzy ontologies concepts denote fuzzy sets, relations denote over classes, instances and properties that characterize their features.
fuzzy relations, and axioms and facts are not in general either true or Descriptions Logics (DLs) provide a formal substratum to ontology
false, but they may hold to some degree of truth. Fuzzy ontologies are representation primitives by mathematically defining the constructors
represented by using fuzzy ontology languages, and can be queried by that can be used to form complex classes, properties, and axioms, as
using fuzzy ontology reasoners, such as DeLorean [14]. Although fuzzy well as their semantics. In particular, the OWL 2 language, the cur-
ontologies have been used in different information science research rent standard for Semantic Web ontologies [9], is based on the DL
areas – e.g., information retrieval [15], knowledge merge and summari- named SROI QðDÞ (each letter corresponds to a constructor or set of
zation [16–19], recommender systems [20,21], and decision-making constructors). A detailed description of DLs is out of the scope of this
[22] – , to the best of our knowledge they have not been yet applied to paper, but the interested reader can found a concise summary in [24].
solve industrial problems in practice. Beetz et al. proposed in [25] a mapping from the IFC data model to
The overarching objective of this paper is to present the fundamen- OWL that generates an ifcOWL ontology. Later works have implemented
tal characteristics and the applications of fuzzy ontologies that can be of procedures to convert a given BIM in STEP format to RDF instances in
interest to the BIM users. Rather than focusing on the formal description order to obtain a specific semantic BIM [26]. The IFC-to-RDF conversion
of the mathematical foundations of fuzzy DLs, we provide examples that software is a publicly available tool that performs both tasks [23]. In this
show the representation and reasoning capabilities of such formalisms. section, we describe some aspects of the models obtained by the IFC-to-
To do so, we extend the ifcOWL and ifcRDF models obtained by the RDF tool that are relevant for our fuzzy extension. Notice that, as men-
IFC-to-RDF conversion tool [23] with fuzzy information. In addition, tioned by the authors, the translation of a model is not unique, since dif-
we describe how they can be exploited in different use cases that illus- ferent conversion strategies can be applied depending on the user
trate common problems in the building design and analysis stages. needs. We will focus on a slightly modified version of the ‘OWL 2 EL–
Accordingly, the main contributions of the paper are the following: RDF List’ ontology.1 To increase readability, we will use the OWL Man-
chester syntax in the following examples [27].
• We provide a description of the main features of fuzzy ontologies in In the conversion, IFC EXPRESS classes are mapped into OWL
the context of the Linked Building Data initiative, avoiding the cum- classes, and subtype and supertype relations are represented with
bersome details of the underlying theoretical framework. For the in- class inclusion axioms. For example, the IfcWindow entity is represented
terested reader, we also provide a selection of pointers to related as follows:
works that elaborate on these topics.
• We present illustrative examples of the main representation primitives
of a typical fuzzy Description Logic, and how they can be applied in dif-
ferent use scenarios. We also explain how to use a fuzzy ontology rea-
soner to query the resulting fuzzy semantic BIM for practical purposes.
• We discuss the advantages of using fuzzy ontologies over non-fuzzy Analogously, attributes are translated into OWL properties. Due to
(crisp) representations in the scope of the Linked Building Data some particular features of EXPRESS, such as the rich data type system
research area, considering the current state of the art and the level of and the capability to define attributes local to classes, the conversion
maturity of the existing tools, as well as their interrelation with the of properties is not straightforward. Among the possible alternatives,
ifcOWL and ifcRDF technologies. the authors of IFC-to-RDF have successfully used property reification,
wrapper classes for data types, and variant names for local properties.
The snippet below represents the overallHeight_of_IfcWindow
The remainder of the paper is structured as follows. First, we describe attribute, which translates into a functional DataProperty property
the materials, methods and tools used in this research work; namely: with defined domain and range. In OWL 2, it would be possible to define
(i) ontologies as representation formalisms for the BIM; (ii) fuzzy logic a range restriction based on a facet to delimit the values allowed for the
and fuzzy DLs for the creation of fuzzy ontologies; (iii) reasoning with attribute:
fuzzy ontologies. Next, we proceed to describe some relevant representa-
tion primitives of the selected fuzzy ontology language. We explain the
meaning of each primitive with examples of their use to represent impre-
cise building data and to make fuzzy queries. We elaborate afterwards on
the added value of such extensions in three specific use cases: soft inte-
gration of cross-domain knowledge, flexible BIM query and imprecise
parametric modeling. We discuss the limitations of fuzzy ontologies and
their implementation feasibility in a BIM context, especially from a per-
formance perspective. Finally, the paper finishes with a summary of the
most important conclusions achieved and some directions for future
work. 1
http://ugritlab.ugr.es/r/ifc/schema-EL-RDFList.owl.
204 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212

Class definitions can include additional restrictions based on proper- 2.2. Fuzzy ontologies
ties besides class inclusion axioms. For instance, following the previous
example, the class definition below states that any IfcWindow is related 2.2.1. Fuzzy Logic
with the property overallHeight_of_IfcWindow only to float Our extension to manage imprecise knowledge in the semantic BIM
values, and at most to one of them: is based on fuzzy Description Logics, a Fuzzy Logic-based extension of
Description Logics. Fuzzy Logic and Fuzzy Sets were proposed by
L. Zadeh [31] to manage imprecise and vague knowledge. While in
classical set theory elements either belong to a set or not, in fuzzy set
theory elements can belong to a set to a certain degree.
More formally, let X be a set of elements called the reference set.
A fuzzy subset A of X is defined by a membership function μA(x), or
simply A(x), which assigns any x ∈ X to a value in the interval of real
numbers between 0 and 1. As in the classical (or crisp) case, 0 means
no membership and 1 full membership, but now a value between 0
Instances are defined by asserting its type and the property and 1 represents the extent to which x can be considered an element
values. This example shows the definition of a window instance of X. Some membership functions commonly used to define fuzzy sets
named window_1: are the trapezoidal and the triangular. Classical examples of concepts
that can be described using fuzzy sets are Tall or Young.
In Fuzzy Logic, it is common to restrict to finite chains of degrees of
truth, instead of using the real interval [0, 1]. In our case, we will work
n o
with the finite chain of p + 1 elements: N ¼ 0 ¼ γ0 b γ 1 b… b γp¼1 ,
where p ≥ 1 [32]. Such N can be understood as a set of linguistic
terms or labels; for example, {0, 0.2, 0.4, 0.6, 0.8, 1} →
{false, closeToFalse, slightlyFalse, slightlyTrue, close
ToTrue, true}. From a practical point of view, a small p is sufficient
in many applications, and preferred against a large p to facilitate the in-
The use of OWL and RDF to represent IFCs allows circumventing
terpretability of the semantics to the users. Fig. 1.a shows an example of
the lack of mathematical formality of EXPRESS. This has two direct
a discretized trapezoidal membership function defined over the set of
consequences. First, the RDF model of the building information can
real numbers, so q1, q2, q3, q4 ∈ ℝ and γ1 ; …; γ p ∈ N .
be shared and linked on the Web of data. We can use the SPARQL lan-
Fuzzy Sets theory extends all classical set operations to fuzzy sets.
guage [28] to formulate flexible and distributed queries that, in gen-
The intersection, union, complement and implication set operations
eral, cannot be easily solved within the IFC standard. Likewise, the
are performed by corresponding functions; respectively, a t-norm, a t-
availability of RDF storage and manipulation tools notably reduces
conorm, a negation, and an implication. The combination of them is
the effort required to implement new products and to support
called a fuzzy logic, and there are several of them depending on the se-
unpredicted users' needs.
lected functions. We will consider the fuzzy connectives originally pro-
Second, existing OWL inference engines can be used to reason with
posed by Zadeh, namely the Gödel conjunction and disjunction,
the semantic BIM models. Reasoning within an ontology is an automatic
Łukasiewicz negation, and Kleene–Dienes implication (Table 1). Other
procedure that infers new knowledge that has not been explicitly in-
typical fuzzy logics are Łukasiewicz, Gödel, and Product, which have dif-
cluded in the ontology but is a logical consequence of the represented
ferent properties [33].
axioms. For instance, a valid inference is that “window_1 is an instance
Relations can also be extended to the fuzzy case. A (binary) fuzzy
of IfcBuildingElement”, because window_1 is an instance of
relation R over two countable sets X and Y is a function R : X  Y→N .
IfcWindow, and IfcWindow is a subclass of IfcBuildingElement. It
Several properties of the relations (such as reflexive, irreflexive,
goes without saying that these inferences become more complicated
symmetric, asymmetric, transitive, or disjointness) and operations
when complex concept expressions are used in large knowledge
(inverse, composition) can be trivially extended to the fuzzy case.
bases. The reasoning algorithms are implemented by reasoning engines,
Fuzzy modifiers (or hedges) apply to fuzzy sets to change their
such as HermiT [29].
membership function. For instance, given a fuzzy set Tall we may
As a matter of example, we can define a complex class and re-
want to define the fuzzy set of very Tall people by using an appropriate
trieve all the (explicit and implicit) instances of that class. The
following expression denotes all the building elements built of modifier. Formally, a modifier is a function f m: N →N . Examples of
concrete: modifiers are very, more or less, and slightly. They can be defined by
using different types of membership functions; e.g., triangular functions
(Fig. 1b).
Eventually, changing the usual true/false convention leads to a new
type of logical propositions, called fuzzy propositions. Each fuzzy prop-
osition may have a degree of truth in N , denoting the compatibility of
the fuzzy proposition with a given state of facts. For example, x is a big
window ≥ 0.8 says that we have a fairly big window (the degree of
truth of x being a big window is at least 0.8).

Reasoning can be also useful to facilitate the detection of incon-

sistencies in the representation, and therefore, to support complex 2.2.2. Fuzzy Description Logics
modeling. Furthermore, it is possible to extend the model with addi- In this work we have considered a fuzzy extension of the SROIQðDÞ
tional functionalities based on other Semantic Web technologies. DL, which in turn corresponds to a fuzzy extension of the OWL 2
This is the case of [30], where logic-based rules for building perfor- language. This fuzzy DL, presented in [34,35], is a subset of the logic
mance checking are defined with notable advantages over more tra- supported by the fuzzy reasoning engine DeLorean used in the next
ditional approaches. section.
 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212 205

Fig. 1. (a) Trapezoidal membership function; (b) triangular fuzzy modifier.

We summarize in Table 2 the main concept and role constructors of of DLs, while those concerning degrees are specific of fuzzy DLs. Usually,
the fuzzy SROI QðDÞ. It can be seen that they have a direct correspon- reasoning tasks can be reduced to f K consistency [36]. The most
dence with their crisp counterparts. The syntax is equivalent to the interesting ones are informally described below. Notice that in all
language accepted by DeLorean, which is a variation of the Manchester cases, the elements that are used as input to the reasoning task may
syntax for OWL 2 in a more functional-programming style. We purpose- not be explicitly included in the ontology.
ly do not include the formal semantics of these expressions, which can
be found in the previously mentioned works. The notation used in • Fuzzy knowledge base consistency (or satisfiability): check if all the
Table 2 is the following: axioms in the knowledge can be satisfied; i.e., they do not contradict.
• Concept satisfiability: check if a given concept does not correspond to
• a, b are individuals. an empty set of instances.
• A is an atomic (simple) concept and C, C1, are (possible complex) • Entailment: check if a given axiom is entailed by the explicit axioms of
concepts. the knowledge base.
• R is a fuzzy role.2 • Concept subsumption: check if there is a subsumption relation
• M is a strictly positive natural number and N is a natural number. between two given concepts.
• d is a fuzzy datatype, defined with the syntax (trapezoidal • Instance retrieval: retrieve all the instances of a given concept
q1 q2 q3 q4) (q1, q2, … are real values, as depicted in Fig. 1). (optionally, with a minimum degree).
• D, D1, are degrees in the finite chain N . • Best degree bound (bdb): get the maximum degree to which an axiom
– In fuzzy nominals, D is in {γ1, …, γp}. If omitted, γp is assumed. holds.
– In axioms, D is in {γ1, …, γp} if preceded by N =, or in {γ0, …, γp − 1} if • Maximal concept satisfiability degree: get the maximum degree to a
preceded by b =. which an individual can belong to a given concept.
• | denotes alternative sub-expressions.
• [ ] denotes optional sub-expressions.
• { } denotes grouping to make precedence explicit. As in the crisp case, reasoning with fuzzy ontologies is performed
with reasoning engines. The two most prominent fuzzy engines are
fuzzyDL [37] and the previously mentioned DeLorean (DEscription
A fuzzy Knowledge Base (f K), or just a fuzzy ontology, is therefore LOgic REasoner with vAgueNess) [14]. Both can be freely used and sup-
composed by a finite set of fuzzy axioms. The axioms are grouped into port expressive fuzzy ontology languages. The main difference between
a fuzzy ABox describing individuals, a fuzzy TBox describing concepts, them is that they apply distinct strategies to carry out the reasoning
and a fuzzy RBox describing roles. Table 3 shows the axiom constructors process. fuzzyDL implements a mixture of tableau algorithms and a
of the fuzzy DL SRI OQðDÞ. Note that the bounds for the degrees are MILP optimization problem, whereas DeLorean implements a reduction
optional, so if N= D is omitted, N= γp is assumed. procedure that transform a fuzzy ontology into a crisp ontology that can
Examples of the use of the most relevant fuzzy representation prim- be processed by any non-fuzzy DL reasoning engine. In terms of efficiency,
itives are offered in Section 3. We do not use all the primitives listed in fuzzyDL includes several optimizations to reduce the time required for
Table 2, but they are included for the sake of completeness. Particularly, the most common reasoning tasks [38], whereas DeLorean exploits the
all, one-of and self are not considered. These would be some typical cases in which recomputing the reduction of the fuzzy ontology is not
(simplified) examples of their use: necessary, and the capability of using different crisp reasoners under
the hood. Given our scenario, in which a non-fuzzy ontology is already
available, using DeLorean for fuzzy ontology reasoning is an appropriate
choice.

3. Results and discussion

3.1. Fuzzy ontology constructors and axioms in the BIM

2.2.3. Reasoning with fuzzy ontologies
In fuzzy DLs there are many reasoning tasks involving the axioms of To illustrate the use of the representation primitives introduced in
the knowledge base. Some of them are extensions of the reasoning tasks Tables 2 and 3 in the construction domain, in this section we describe
some examples based on the model files provided with the IFC-to-RDF
2
Strictly speaking, there are certain syntactic restrictions in some expressions regarding tool. To facilitate the reproduction of the examples while maintaining
the use of roles, but we omit here the details. simplicity, we have used a sample IFC file that describes a model of a
206 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212

Table 1 Furthermore, we can use more complex concept expressions. For

Specification of the Zadeh fuzzy logic. example, we can ask the reasoner to retrieve all the instances of the
Notation Function intersection class IfcMaterial and IfcMaterialLayer. Let us sup-
t-norm α⊗β min{α, β}
pose that we have also asserted:
t-conorm α⊕β max{α, β}
negation ⊖α 1−α
implication α⇒β max{1 − α, β}
Therefore, only material_2 would be inferred as an instance of the
intersection class. The degree of this result would be the (Gödel)
conjuction of the degrees of belonging to IfcMaterial (0.8) and to
IfcMaterialLayer (0.6); i.e., min(0.8, 0.6) = 0.6.
wall with an opening section corresponding to a window. Notice how-
ever that the presented examples are valid in any model regardless of 3.1.2. Fuzzy role assertions
its size and complexity, since constructors and axioms are applied Fuzzy roles denote fuzzy relations between ontology instances.
exactly in the same way. Detailed use cases in a wider BIM context are A fuzzy role assertion materializes an association between two in-
described in Section 3.2. stances that holds to a degree. To create a fuzzy role assertion, we
The ifcOWL and the ifcRDF files resulting from the conversion with use the axiom A2. Similarly, we can use axiom A3 to denote that
the ‘OWL 2 EL–RDF List’ options have been slightly adapted to make two instances are not related to a degree. same – as and
the explanations more readable. The sample model, which is depicted different – to are special relations represented with the axioms A4
in Fig. 2, can be also obtained from the authors' web page.3 The fuzzy and A5, respectively. Typical fuzzy DLs, and particularly our fuzzy S
extension has been created with the Protégé [39] ontology editor and ROI QðDÞ, do not fuzzify these two relations. Therefore, if we want
the Fuzzy OWL 2 language plugin [40] which supports exporting to to model that two individuals are similar, usually a new similarity
the native format of the DeLorean reasoner. The complete process is relation is defined.
shown in Fig. 3. The interested reader can find more details on the
practical use of DeLorean in [14]. Example 2. A new fuzzy role has been defined in the ontology to relate
In the remainder of the section, we will use the DeLorean syntax. We the similarity degree between two building materials, namely the
will note the terms defined in the original ifcOWL ontology with the similar_to_IfcMaterial object property. This property can be de-
prefix ifc, and our additions with an empty prefix (:). The examples fined as symmetric (R9), because it holds in both directions (with the
are built incrementally from the initial model in such a way that, if not same degree), and transitive (R8). By extension, it would be possible
explicitly stated, the additions in former examples apply in later exam- to define other features of the property with the axioms R3–R14: reflex-
ples. The final file with the fuzzy ontology in DeLorean format, including ive, irreflexive, functional, etc. Let us also suppose that we have in the
all the examples, can be also found at the authors' web page.4 fuzzy ontology additional instances of IfcMaterial representing
‘mortar’ and ‘ecologic mortar’ materials. We can now assert that ‘con-
crete’ is quite similar to ‘mortar’, but ‘mortar’ is only moderately similar
3.1.1. Fuzzy concept assertions to ‘ecologic mortar’.
As already mentioned, in a crisp ontology, instances either belong to
a class or not. In a fuzzy ontology, it is possible to assert that an instance
belongs to a class to a degree. This degree can be an unrestricted real
value in [0, 1], or as in our case, a value in a finite chain N . This kind
of axiom is called a fuzzy concept assertion, and is noted with
instance, as shown in Table 3 (A1). We will assume a finite chain
N = {0, 0.2, 0.4, 0.6, 0.8, 1}.

Example 1. If we consider the class IfcMaterial as a fuzzy concept,

we can add a new instance representing “paper” that can be only par-
tially considered a material. (Note that IfcMaterial already has an in-
dividual, material_1, as depicted in Fig. 2.)

From this point on, we can retrieve those instances that can be con- A possible query would be to retrieve the materials that are quite
sidered materials at least to a degree. Following the example, we can ask similar to “concrete”; i.e., those belonging to the following class with
the reasoner to query all the instances of IfcMaterial with degree degree ≥ 0.6:
≥ 0.6, which will trivially include material_1 (degree 1) and
material_2 (degree 0.8). Many other fuzzy concept assertions could
be added to denote elements that, for any reason, we do not want to
fully belong to a type, or we do not know exactly: additional materials,
spaces, equipment, etc.

Since the similar_to_IfcMaterial property is symmetric

and transitive, the query returns the results material_1 (1),
material_3 (0.8) and material_4 (0.6 = min(0.8, 0.6), due to
3
http://ugritlab.ugr.es/r/ifc/sample-model-EL-RDFList.ttl. the propagation of the transitivity with the Gödel conjunction
4
http://ugritlab.ugr.es/r/ifc/sample-model-DeLorean.dlr. operation).
 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212 207

Table 2
Concept and role expressions in fuzzy SROI QðDÞ.

However, a large amount of building materials are part natural, be-

cause their raw components, and part artificial, because they are proc-
essed, refined, or mixed with other components. This context can
modeled with an imprecise taxonomy created by using fuzzy GCIs. For
example, we can consider that ‘plywood’ is, to some extent, a vegetal
material, because it is mainly made of wood, but also an artificial mate-
rial, because it is glued with chemical products. Similar fuzzy GCIs can
An extension of this query would be to obtain all the building be used to define materials like ‘glass’ or ‘reinforced concrete’:
elements that are built with materials somehow similar to ‘mortar’.
These individuals would be the instances of the following complex
class filtered by degree ≥ 0.2:

Based on this fuzzy taxonomy, we can assign our material instances

to classes by using (fuzzy) concept assertions (replacing those defined
The result of the query would be wall_1 (0.8), but we can imagine in the original model and Examples 1–2):
that in a larger model it would give less evident outputs.

3.1.3. Fuzzy general concept inclusions

General concept inclusion axioms (GCIs) in crisp DLs, noted
C 1 ⊑ C 2 , represent the notion that any instance of the class C 1 is
also an instance of class C2. GCIs are often read as C 2 subsumes C 1. Afterwards, we can query the model to retrieve
A fuzzy GCI, represented with the implies-concept axiom (C1), ArtificialMaterials with degree ≥ 1. The results would be:
states that C1 is a subconcept of C2 to degree at least D. As expect- material_1, material_3, and material_4 (degree propagation is
ed, C1 and C2 may be complex concepts; i.e., concepts denoted by calculated with the Kleene–Dienes implication function). Note that
complex expressions built with the constructors enumerated in material_5 is not in the result list because it belongs to
Table 2. Based on the idea of GCI, the logic also allows other restric- ArtificialMaterial only with degree 0.8.
tions such as concept equivalency (C2) and disjointness (C3). The More interestingly, we can obtain all the elements of the building built
corresponding primitive of GCIs for roles are role subsumption ax- with artificial materials by extending the query concept; in our case,
ioms (R1), which can use chains of roles in the right part of the wall_1:
subsumption.

Example 3. Let us suppose that a taxonomy of materials is added to

the model. This taxonomy includes several material types divided in
two main categories, artificial and natural, which establish a non-
strict partition of IfcMaterial.
208 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212

Table 3
Axioms in fuzzy SROI QðDÞ.

−
−
−

−
−
−
−

−
−

−
−

−
−

Moreover, we can query the elements built with materials similar to “high” is characterized by a trapezoidal function, which will calculate
artificial materials (in our case, wall_1) by reusing the fuzzy similarity the “degree of being high” for any given real value. Similarly, we use
property defined in Section 3.1.2: another trapezoidal function to characterize the degree of being wide
for any given real value:

Trivially, the queries can be adapted to other material types, and by

extension, to any other fuzzy taxonomy integrated into the model. It is
worth to highlight than fuzzy taxonomies can coexist with crisp taxon-
omies, thus providing high representation flexibility while avoiding the
user to change the whole model.

3.1.4. Fuzzy datatypes Now, it is possible to query all the big windows of the model,
Fuzzy datatypes are the natural extension of crisp datatypes, since being them those instances of the class (and High_IfcWindow
they allow imprecise statements over a concrete domain. The truth Wide_IfcWindow). Since in the sample file the window size is
value of a datatype predicate is given by a (discretized) function, (width × height) = (1.0 × 1.5), issuing such query to the reasoner
which is typically a trapezoidal function like the one depicted in would give as a result window_1 (with degree 0.4 = min{0.4, 0.6}). To
Fig. 1a. Fuzzy data types can be used in several concept expressions, as obtain this result, we use the Gödel conjunction to apply the and
noted in Table 2 with d. operator over the membership values calculated with the trapezoidal
Let us define two new classes based on fuzzy datatypes: a “high fuzzy datatype (Fig. 4). If necessary, these values would be discretized
height window” and a “wide width window”. A high height window to fit the ones in N . We can easily modify the query to work with
is a window that has some “high” height value. The fuzzy notion of building elements other than windows, or to add more conditions to
“high” is characterized by a trapezoidal function, which will calculate restrict ourselves to elements located inside a certain level, having a
the “degree of being high” for any given real value. Similarly, we use relation to other elements, with a given property, etc.
another trapezoidal function to characterize the degree of being wide
for any given real value:
3.1.5. Fuzzy modifiers
Example 4. Let us define two new classes based on fuzzy datatypes: a Fuzzy modifiers are used to change the meaning of a fuzzy concept
“high height window” and a “wide width window”. A high height win- by modulating its membership function. Roughly, a fuzzy modified con-
dow is a window that has some “high” height value. The fuzzy notion of cept is therefore a variation of the original concept that assigns slightly
 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212 209

Fig. 2. Fragment of the baseline model.

different membership degrees to its instances. Our fuzzy SROI QðDÞ information modeling and retrieval. The fuzzy ontology framework
allows the use of the fuzzy modifier triangular, defined by a function precisely defines the semantics of these components and how they
like the one presented in Fig. 1b. can be combined, letting the users to use them at their convenience.
In this section, we elaborate on different use cases that show how the
Example 5. Based on the class High_IfcWindow created in Example 4, fuzzy semantic BIM can address common problems appearing during
we can define the class of windows with very high height with the the building life-cycle that cannot be solved with a non-fuzzy BIM.
expression:
3.2.1. Cross-domain knowledge linking
Fuzzy general concept inclusions (Section 3.1.3) are the backbone of
a fuzzy ontology. Similarly to the crisp case, fuzzy GCIs trigger the most
In consequence, window_1 is a Very_High_IfcWindow with de- interesting inferences, because we can use complex concept construc-
gree 0.2, resulting from the modification and discretization to N value tors in the left and the right part of these axioms (e.g. Example 5). Essen-
of the High_Ifc membership value with the triangular function tially, fuzzy GCIs allow us to define imprecise concept inclusions, in such
(Fig. 5). Naturally, it is also possible to define more complex concepts; a way that we can quantify the degree of overlap between two classes
e.g., to select building spaces with very high windows. of individuals, and then conveniently operate with this degree within
the Fuzzy Logic framework. As in the crisp case, fuzzy GCIs can be
3.2. Use cases for the fuzzy semantic BIM multi-dimensional, in the sense that multiple hierarchies can be defined
relating the same concepts.
The previous section describes the basic building blocks of a fuzzy One direct application of fuzzy GCIs is the integration of heteroge-
ontology, and explains how they can be used for imprecise building neous building entity taxonomies. Let us suppose that we have two

Fig. 3. Processing workflow from the original files to the fuzzy queries.
210 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212

a) b)

Fig. 4. (a) Fuzzy datatype for High_IfcWindow; (b) fuzzy datatype for Wide_IfcWindow.

different catalogues of materials supplied by two different contractors. imprecise definitions, which are more appropriate in several cases. For
Usually, we cannot expect that both catalogues will use the same instance, a concept like room occupancy, which is important for energy
names and codes for the material types. Moreover, it may happen that facility managers, is better expressed with fuzzy values (high, normal,
the material types in the catalogues do not directly correspond to the etc.), similar to the ones described in Section 3.1.4. From this representa-
ones used in the BIM. This situation requires developing a mapping tion, it is possible to retrieve information from the BIM and perform addi-
between the taxonomies to establish a semantic link between related tional calculations by aggregating data; e.g., to calculate the approximated
concepts. The mapping would support to automatize calculations such expected energy consumption from the specification of individual room
as the construction cost of a part of the modeled building using mate- equipment, load predictions, and occupancy profiles. Interestingly
rials from a selected contractor—this can be implemented based on an enough, all the parameters of this kind of 'what if' scenarios do not
ontology query that returns the price of the materials matching the need to be exactly assigned. Other scenario is the classification of build-
ones in the BIM, similar to the one shown at the end of Section 2.1, ings or building components according to their features. For example,
and then multiplies them by their area of use. In the crisp setup, the we can first define imprecise end-user concepts from IFC elements (a
mappings between material types are binary (yes/no), whereas with breezeway, a dining room), and then create a flexible building description
the fuzzy BIM, it is possible to establish a degree of similarity between (a dogtrot house typically has one story, a breezeway, and at least two
material types (as in Example 3). Therefore, with our proposal, the rooms about 20 ft wide) readily-available for BIM data retrieval and
query can be expanded to material types similar to the ones actually exchange.
selected in the BIM at least to a degree, or even implement a cost calcu-
lation that increases the price of a piece of material depending on this 3.2.2. Imprecise BIM queries
similarity value. Naturally, this approach can be extended to map In the explanation above, we have implicitly assumed that the
other building element classifications; for example, those defined in creation of concept definitions makes it possible to retrieve the
the International Building Code [41]. instances of these new concepts. That means that, for example, if we
From a broader perspective, the fuzzy semantic BIM increases define the (fuzzy) concept Dogtrot House, we can query a semantic
the knowledge integration features provided by the Semantic Web BIM by using a (fuzzy) reasoning engine to retrieve the instances of
technologies, thus improving interoperation, a typical problem in the this concept; i.e., if there is only one building represented in the
architecture, engineering and construction industry [42]. In general, model, the building instance itself (with a fulfillment degree) or an
cross-domain and cross-cultural knowledge can be more flexibly incor- empty set. In this section, we study in more detail the query capabilities
porated and managed in the building knowledge base. Following the of the fuzzy semantic BIM, which provides an open, formal language for
previous example, we can see that compound domain-specific fuzzy the definition of imprecise sophisticated queries, relaxed constraint
concepts can be now formally defined and linked. One application checks, and partial model filters.
scenario of this is creating a vocabulary to facilitate the exchange of infor- In the literature, we can find some proposals aimed at the creation of
mation between the stakeholders involved in the construction process non-proprietary BIM query languages. ifcRDF and ifcOWL are contribu-
(designers, constructors, facility managers, construction workers, etc.) tions in that regard, since the resulting models can be loaded in a
Some of these actors are not expected to use the IFC conventions, and triplestore or a reasoning engine, and handled with standard SPARQL
therefore, it can be very helpful to formally define their daily-use and OWL queries. Similarly, BIMQL is a language for generic querying
concepts in standard terms. The fuzzy approach allow us to work with of IFC-based BIM models [43], supporting free-variable queries and
model updates, among other features. Our fuzzy ontology model
extends the capabilities of these approaches by allowing imprecise
knowledge retrieval. A straightforward example of a fuzzy query has
been presented in Example 4, in which we define a fuzzy concept
based on a fuzzy datatype value, and then retrieve the model instances
belonging to this concept, as well as their membership degree. In
general, fuzzy role assertions and fuzzy datatypes enable storing
imprecise information (e.g. a fuzzy dimension value), and formulating
imprecise queries over precise and imprecise data (e.g. retrieve
elements with dimension values within an imprecise range). This
kind of queries can be used to obtain a filtered view of specific
building elements, with the advantage that the condition is more
flexible and the results are ordered by the degree of fulfillment of the
condition.
Fuzzy role assertions can be also applied to create imprecise geomet-
Fig. 5. Fuzzy modifier for Very_High_IfcWindow. ric properties. For instance, it may be interesting to define a fuzzy
 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212 211

property representing how close two arbitrary building elements are, 3.3. Discussion
similar to the one in Example 2. The degree of such relation can be
automatically computed with a distance measure from the building The previous examples and use cases show that the proposed fuzzy
geometry, and then stored in the model as fuzzy data value. Notice extension notably increases the capabilities of the semantic BIM, and
that this would require calculating pair-wise distances between all provides support for much more expressive representations. Queries
building elements, which may be computationally expensive and retrieve not only data explicitly asserted in the model, but also informa-
require additional optimizations. Other interesting improvement can tion automatically inferred by a logic-based reasoning process.
be the combination with RCC (Region Connection Calculus) predicates, Fuzzy DLs offer a sound theoretical framework for such fuzzy ontol-
which allow symbolic representation and reasoning with topological re- ogies, and the availability of languages and tools facilitates the
lations [44]. RCC is not directly supported in crisp OWL, but still, it is implementation of this kind of solutions. We have used a fuzzy ver-
possible to create axioms to model common relations, like tangent, sion of the SROI QðDÞ DL, but it would be possible to change the un-
overlap, disjoint, etc., and instantiate them by relying on an external derlying logic without any loss of generality. Actually, other even
module [45]. By extension, fuzzy ontology axioms can be used to more expressive logics, including supplementary representation
formally define soft positioning constraints; e.g., small overlapping, ap- primitives targeted at particular needs of BIM users could be consid-
proximately tangent, etc. If basic fuzzy axioms are combined to other ered. Some of them have been mentioned through the paper: spatial
fuzzy property and concept constructors, we can represent for instance reasoning, alternative families of fuzzy logic operators, constraint
that there are several objects in the surroundings of a big element. If checking, etc.
these axioms are contradictory to others in the BIM, the reasoning However, adding additional representation layers to the semantic
engine will infer that the model is inconsistent, which can be useful BIM comes with a cost. Expressive ontologies and fuzzy ontologies are
for the detection of clashes. In contrast to the crisp approach, in the known by their inferior performance, which can be a serious drawback
fuzzy case we can impose a threshold degree, and therefore relax when working with larger models. To test the scalability of our proposal,
these restrictions at convenience. Last but not least, specific extensions we have executed the same example queries of Section 3.1 with the du-
of fuzzy DLs could be exploited to natively support fuzzy geometric plex apartment model used in the 2009 COBie Challenge.5 These tests
reasoning in the BIM [46]. have proved that efficiency quickly degrades with such realistic models,
even to the point of making them unusable because of the high time
3.2.3. Fuzzy parametric modeling needed to solve relatively simple queries (over 1 min for Examples 4
The notion of soft positioning constraint presented before is related and 5). We have to mention, however, that no specific optimization
to that of parametric modeling in normal BIMs. Parametric models are strategies were applied to the fuzzy inference procedure.
based on parametric components, which are virtual BIM elements that One possible solution to this issue would be to restrict the semantic
have associated a range of possible values to properties (a numeric BIM to a tractable OWL profile. As a matter of fact, this was the purpose
interval for a dimension, a set of colors for a solid element) rather of using the EL strategy of the IFC-to-RDF tool, which produces a slightly
than a fixed value. Accordingly, parametric components are usually less expressive but more efficient model. Due to its internal design, if the
provided as templates, sometimes by third-parties, to be instantiated fuzzy ontology used in DeLorean is based on OWL 2 EL, a DL engine op-
in the BIM. This allows the designer to reuse components, and even timized for this subset of OWL can be used in the last reasoning step
to place parametric components in the model and let the software to (e.g., ELK [48]). Nevertheless, the output produced by IFC-to-RDF is
optimize the parameter values according to the requirements imposed not a strict OWL 2 EL ontology (e.g., class unions are generated). This
by the non-parametric components. The latter requires solving an also indicates that a more precise characterization of the conversion
optimization problem expressed as a system of linear inequalities. procedure could be useful to make better use of the features of available
As introduced in the previous section, some kinds of fuzzy axioms techniques and tools.
express semantics similar to parametric constraints. In particular, a Other solutions specifically aimed at improving the performance of
fuzzy role is not very different to a parametric property: it also defines the reasoner itself would be to pre-calculate some common queries
a plausible value range, although all the values may not totally satisfy when a model is firstly created, such as the concept hierarchy or the
the property. In this regard, we can rely on concepts similar to the membership values of selected fuzzy datatypes. Since we may expect
ones in Example 4 to define fuzzy axioms establishing that a building several queries concerning the geometric properties of building
story must include a big room and two small rooms, and the small elements, the latter may have a notable impact in the efficiency. In addi-
rooms must be very close. Given an instantiation of these concepts and tion, DeLorean can pre-calculate the reduction from the fuzzy ontology
relations, via crisp or fuzzy position and dimension values, we can test to the crisp ontology, which remains unchanged if no new instances
whether the constraints are satisfied or not by testing if the model is are added (as it may be expected after the initial model creation
consistent. Similarly, finding a model that maximizes the degree of ful- stage). This is consistent with the foreseen use of our proposal. As it
fillment of the fuzzy axioms representing the constraints roughly corre- can be concluded from the examples, the fuzzy extension is likely to
sponds to finding the best degree bound of these axioms. The fuzzy be more useful in the analysis and decision stages of a project.
approach has important advantages compared to the crisp approach,
because relaxing the constraint to a degree lesser than 1 allows reducing
the problem of conflicting parameters in overconstrained models. 4. Conclusions and future work
Unfortunately, only a limited set of constraints can be modeled with
fuzzy axioms, and in the best case, they require intensive geometric In this paper, we have presented a fuzzy logic-based extension of
calculations that are not directly performed within the ontology. For in- the semantic BIM proposed within the Building Linked Data research
stance, in the previous example we cannot find the best room arrange- initiative. Our proposal allows imprecise data linking and querying
ment among all the possible positioning alternatives by only using the to ontological building models within the well-defined theoretical
fuzzy reasoner. Therefore, we consider that the fuzzy BIM can be helpful framework of fuzzy Description Logics. The resulting models support
to encode imprecise constraints, but solving the optimization problem very expressive imprecise queries, which offer new ways to retrieve
would require the implementation of an external fuzzy constraint information not available in the current systems. Unfortunately, fuzzy
satisfaction algorithm [47]. Nevertheless, it is worth to mention that inference engines are experimental tools, and therefore, they present
the fuzzyDL reasoner provides limited support for systems of fuzzy some drawbacks in terms of efficiency (high execution time) and ease
inequalities. Exploring these capabilities is an interesting direction for
future work. 5
http://www.nibs.org/?page=bsa_commonbimfiles.
212 J. Gómez-Romero et al. / Automation in Construction 57 (2015) 202–212

of use (the only supporting tool to facilitate the creation of a fuzzy [18] R.-C. Chen, C.T. Bau, C.-J. Yeh, Merging domain ontologies based on the WordNet
system and fuzzy formal concept analysis techniques, Appl. Soft Comput. 11 (2)
model is the Fuzzy OWL Protégé plugin and API [14], which is not (2011) 1908–1923.
suitable for medium-scale ontologies). However, the initial results are [19] K. Todorov, C. Hudelot, A. Popescu, P. Geibel, Fuzzy ontology alignment using back-
promising and useful in different scenarios, and several optimizations ground knowledge, Int. J. Uncertainty Fuzziness Knowledge Based Syst. 22 (1)
(2014) 75–112.
can be applied in the near future. [20] C.-S. Lee, M.H. Wang, H. Hagras, A type-2 fuzzy ontology and its application to
Besides these improvements, a prospective research direction is to personal diabetic-diet recommendation, IEEE Trans. Fuzzy Syst. 18 (2) (2010)
continue the development of more complex use cases with the collabo- 374–395.
[21] J.A. Rodger, A fuzzy linguistic ontology payoff method for aerospace real options val-
ration of BIM users. This will help to determine the most appropriate uation, Expert Systems with Applications 40 (8) (2013) 2828–2840.
fuzzy DL to be used, and to clarify the role of the fuzzy queries within [22] I.J. Pérez, R. Wikström, J. Mezei, C. Carlsson, E. Herrera-Viedma, A new consensus
the users' workflow. Necessarily, this will require the implementation model for group decision making using fuzzy ontology, Soft. Comput. 17 (9)
(2013) 1617–1627.
of appropriate interfaces to facilitate the interaction with the system
[23] S. Törma, N. Hoang, P. Pauwels, IFC-to-RDF conversion tool (Feb. 2014) [cited Octo-
to users without previous experience in ontological modeling. ber 2014]. URL http://linkedbuildingdata.net/tools/tool-ifc-to-rdf-conversion-tool/.
[24] M. Krötzsch, F. Simančík, I. Horrocks, Description logics, IEEE Intell. Syst. 29 (1)
Acknowledgments (2014) 12–19.
[25] J. Beetz, J. van Leewen, B. de Vries, IfcOWL: a case of transforming EXPRESS schemas
into ontologies, Artif. Intell. Eng. Des. Anal. Manuf. 23 (2009) 89–101.
This research work has been partially funded by the Spanish Minis- [26] P. Pauwels, D. van Derusen, IFC/RDF: adaptation, aggregation and enrichment,
try of Economy and Competitiveness under project TIN2012-30939, Proceedings of the 1st International Workshop on Linked Data in Architecture and
Construction (LDAC 2012), 2012 (Ghent, Belgium).
and by the European Union in the 7th Framework Programme under [27] M. Horridge, P.F. Patel-Schneider, OWL 2 Web Ontology Language Manchester
project Energy IN TIME EeB.NMP.2013-4 (grant agreement no. Syntax, 2nd ed., Dec. 2012. (URL http://www.w3.org/TR/owl2-manchester-syntax/).
608981). The authors have used the IFC-to-RDF tool created by [28] S. Harris, A. Seaborne, SPARQL 1.1 Query Language, http://www.w3.org/TR/
sparql11-query/, Mar. 2013.
Seppo Törmä and Nam Vu Hoang, and the Protégé resource, which is [29] R. Shearer, B. Motik, I. Horrocks, HermiT: a highly-efficient OWL reasoner, Proceed-
supported by grant GM10331601 from the National Institute of General ings of the 5th OWL: Experiences and Directions Workshop (OWLED 2008), 2008.
Medical Sciences of the United States National Institutes of Health. [30] P. Pauwels, D.V. Deursen, R. Verstraeten, J.D. Roo, R.D. Meyer, R.V. de Walle, J.V.
Campenhout, A semantic rule checking environment for building performance
checking, Autom. Constr. 20 (5) (2011) 506–518.
References [31] L.A. Zadeh, Fuzzy sets, Inf. Control. 8 (1965) 338–353.
[32] G. Mayor, J. Torrens, On a class of operators for expert systems, Int. J. Intell. Syst. 8
[1] K. Barlish, K. Sullivan, How to measure the benefits of BIM—a case study approach, (7) (1993) 771–778.
Autom. Constr. 24 (2012) 149–159. [33] G.J. Klir, B. Yuan, Fuzzy Sets and Fuzzy Logic: Theory and Applications, Prentice-Hall,
[2] National BIM Standard-United States, Frequently asked questions about the National 1995.
BIM Standard-United States [cited October 2014], http://www.nationalbimstandard. [34] F. Bobillo, M. Delgado, J. Gómez-Romero, Crisp representations and reasoning for
org. fuzzy ontologies, Int. J. Uncertainty Fuzziness Knowledge Based Syst. 17 (4)
[3] T. Liebich, Y. Adachi, J. Forester, J. Hyvarinen, S. Richter, T. Chipman, M. Weise, J. Wix, (2009) 501–530.
Industry foundation classes release 4 (IFC4) [cited October 2014], http://www. [35] F. Bobillo, M. Delgado, J. Gómez-Romero, U. Straccia, Fuzzy description logics under
buildingsmart-tech.org/ifc/IFC4/final/html/index.htm. Gödel semantics, Int. J. Approx. Reason. 50 (3) (2009) 494–514.
[4] F. Abanda, J. Tah, R. Keivani, Trends in built environment Semantic Web [36] U. Straccia, Reasoning within fuzzy Description Logics, J. Artif. Intell. Res. 14 (2001)
applications: where are we today? Expert Syst. Appl. 40 (2013) 5563–5577. 137–166.
[5] P. Hitzler, M. Krotzsch, S. Rudolph, Foundations of Semantic Web technologies, CRC [37] F. Bobillo, U. Straccia, FuzzyDL: an expressive fuzzy description logic reasoner, FUZZ-
Press, 2011. IEEE 2008, pp. 923–930 URL http://webdiis.unizar.es/~fbobillo/fuzzyDL.
[6] S. Törmä, Semantic linking of Building Information Models, Proceedings of the IEEE [38] F. Bobillo, U. Straccia, Optimising fuzzy Description Logic reasoners with general
7th International Conference on Semantic Computing (ICSC 2013) 2013, concept inclusions absorption, Fuzzy Sets Syst. (2015) (in press).
pp. 412–419. [39] Protégé 5.0 ontology editor and framework (Jun. 2014). URL http://protege.
[7] P. Pauwels, Supporting decision-making in the building life-cycle using Linked stanford.edu/.
Building Data, Buildings 4 (3) (2014) 549–579. [40] F. Bobillo, U. Straccia, Fuzzy ontology representation using OWL 2, Int. J. Approx.
[8] E. Curry, J. O'Donnell, E. Corry, S. Hasan, M. Keane, S. O'Riain, Linking building data in Reason. 52 (7) (2011) 1073–1094 URL http://protegewiki.stanford.edu/wiki/
the cloud: integrating cross-domain building data using linked data, Adv. Eng. FuzzyOWL2.
Inform. 27 (2) (2013) 206–219. [41] International Code Council, International building code, online, 2012.
[9] W3C OWL Working Group, OWL 2 Overview, http://www.w3.org/TR/owl2-overview/, [42] O. Akin, Building Information Modeling: BIM in current and future practice, Ch. Ne-
Dec. 2012. cessity of cognitive modeling in BIM's future, Wiley, 2014. 17–27.
[10] P.J. Hayes, P.F. Patel-Schneider, RDF 1.1 Semantics, http://www.w3.org/TR/rdf11-mt/, [43] W. Mazairac, J. Beetz, BIMQL — an open query language for building information
Feb. 2014. models, Adv. Eng. Inform. 27 (4) (2013) 444–456.
[11] F. Baader, D. Calvanese, D.L. McGuiness, D. Nardi, P.F. Patel-Schneider, The [44] R. Grütter, B. Bauer-Messmer, Towards spatial reasoning in the Semantic Web: a
Description Logic handbook, 2nd ed. Cambridge University Press, 2010. hybrid knowledge representation system architecture, in: S. Fabrikant, M.
[12] E. Sanchez (Ed.), Fuzzy Logic and the Semantic Web, Vol. 1 of Capturing Intelligence, Wachowicz (Eds.),The European Information Society — Leading the Way With
Elsevier Science, 2006. Geo-information 2007, pp. 349–364.
[13] T. Lukasiewicz, U. Straccia, Managing uncertainty and vagueness in Description [45] M. Serrano, J. Gómez-Romero, M. Patricio, J. García, J. Molina, Topological properties
Logics for the Semantic Web, J. Web Semant. 6 (4) (2008) 291–308. in ontology-based applications, Proceedings of the 11th International Conference on
[14] F. Bobillo, M. Delgado, J. Gómez-Romero, DeLorean: a reasoner for fuzzy OWL 2, Intelligent Systems and Applications (ISDA 2011), 2011 (Córdoba, Spain).
Expert Systems with Applications 39 (2012) 258–272 URL http://webdiis.unizar. [46] U. Straccia, Towards spatial reasoning in Fuzzy Description Logics, Proceedings
es/~fbobillo/delorean. of the IEEE International Conference on Fuzzy Systems (FUZZ-IEEE 2009), 2009
[15] T. Andreasen, H. Bulskov, Conceptual querying through ontologies, Fuzzy Sets Syst. (Jeju, Korea).
160 (15) (2009) 2159–2172. [47] M. Delgado, J. Verdegay, M. Vila, A general model for fuzzy linear programming,
[16] C.-S. Lee, Z.-W. Jian, L.-K. Huang, A fuzzy ontology and its application to news Fuzzy Sets Syst. 29 (1) (1989) 21–29.
summarization, IEEE Trans. Syst. Man Cybern. B 35 (5) (2005) 859–880. [48] Y. Kazakov, M. Krötzsch, F. Simančík, The incredible ELK, J. Autom. Reason. 53 (1)
[17] A. Ragone, U. Straccia, F. Bobillo, T.D. Noia, E.D. Sciascio, Fuzzy bilateral matchmaking (2014) 1–61.
in e-marketplaces, Proceedings of the 12th International Conference on Knowledge-
Based and Intelligent Information & Engineering Systems (KES 2008), Part III, Vol.
5179 of Lecture Notes in Computer Science, Springer-Verlag 2008, pp. 293–301.