=Paper=
{{Paper
|id=Vol-2206/paper6
|storemode=property
|title=Concept Learning in Engineering based on Refinement Operator
|pdfUrl=https://ceur-ws.org/Vol-2206/paper6.pdf
|volume=Vol-2206
|authors=Yingbing Hua,Björn Hein
|dblpUrl=https://dblp.org/rec/conf/ilp/HuaH18
}}
==Concept Learning in Engineering based on Refinement Operator==
https://ceur-ws.org/Vol-2206/paper6.pdf
Concept Learning in Engineering based on
Refinement Operator⋆
Yingbing Hua[0000−0002−6857−5586] and Björn Hein[0000−0001−9569−5201]
Faculty of Informatics, Karlsruhe Institute of Technology
76131 Karlsruhe, Germany
{yingbing.hua|bjoern.hein}@kit.edu
Abstract. Semantic interoperability has been acknowledged as a chal-
lenge for industrial engineering due to the heterogeneity of data models
in the involved software tools. In this paper, we show how to learn declar-
ative class definitions of engineering objects in the XML-based data for-
mat AutomationML (AML). Specifically, we transform AML document
to the description logic OWL 2 DL and use the DL-Learner framework
to learn the concepts of named classes. Moreover, we extend the ALC
refinement operator in DL-Learner by exploiting the syntax specification
of AML and show significant better learning performance.
Keywords: Concept Learning · Description Logics · AutomationML
1 Introduction
Engineering is referred to as the activities for designing, testing and commission-
ing complex industrial plants [3]. The engineering life cycle requires efficient data
exchange between software tools, where semantic interoperability plays a central
role. A lot of standardization groups are working at the semantic unification in
various engineering subfields, but a ”Super Data Model” that meets the needs
of all tasks would not appear in the near future [2]. To enable data exchange be-
tween engineering tools, the international standard AML1 (IEC 62714) proposes
an XML-based approach [5]:
– Firstly, AML employs the XML schema from CAEX (IEC 62424) to define
the syntax of engineering data, including classes, attributes, data objects
and their relationships. We call this schema as the abstract conceptual model
of AML.
– Secondly, fundamental engineering concepts are standardized as AML role
and interface classes according to the schema. Role classes define the seman-
tics of engineering objects e.g. Robot. Interface classes define the semantics
⋆
The authors acknowledge support by the European Unions Horizon 2020 research
and innovation programm under grant agreement No 688117 Safe human-robot in-
teraction in logistic applications for highly flexible warehouses (SafeLog).
1
https://www.automationml.org
of interfaces e.g. SignalInterface. They are modeled in subsumption hierar-
chies and can be extended to cover tool-specific terminologies. We call the
role and interface classes as the concrete conceptual model of AML.
– Finally, AML specifies how to use the abstract and the concrete conceptual
model to exchange data between engineering software tools.
The motivation of AML is to neutralize tool-specific data using the afore-
mentioned conceptual models. Nevertheless, the standardized role and interface
classes can not satisfy individual modeling requirements, since they lack of suffi-
cient semantic expressiveness for describing tool-specific engineering objects. In
practice, the user has to extend these classes and leave the semantic interpreta-
tion as a functionality of the data importer. This problem encourages the study
of concept learning from engineering data in AML [6], i.e. inducing descriptions
of named engineering classes in terms of AML class-, attribute- and interface
names.
Semantic web technologies such as RDF(S) and OWL are popular tools to
achieve interoperability in the World Wide Web. Several studies from the engi-
neering domain have shown that transforming XML-based data to RDF(S)/OWL
ontologies enhances the semantic interoperability with automated reasoning [1][8].
In this paper, we transform AML to the description logic OWL 2 DL and de-
rive descriptions of named engineering classes using the well-known framework
DL-Learner [4]. DL-Learner is an open-source project for concept learning in
description logics. Among other important features, a downward refinement op-
erator ρ for ALC was proposed, which was extended to an operator for ALCQ
with the support of concrete roles [10]. Based on ρ, two algorithms OCEL [10]
and CELOE [9] are implemented to learn concepts in description logics. It is
worth noting that DL-Learner employs a partial closed-world reasoner for two
reasons: a) it is much faster than a standard OWL reasoner, and b) machine
learning tasks often desire closed world reasoning. In this paper, we study the
performance of DL-Learner and argue that for our particular task, the refine-
ment operator ρ is not quite efficient since it disregards the constraints exposed
in the underlying XML schema of AML. Therefore, we propose an extension to
ρ in its ALC part in section 3 and compare the results with the original one in
section 4.
2 Preliminaries
Figure 1 shows the main components of AML. Modeling in AML usually begins
with the hierarchy of user-specific roles and interfaces. Subsumption relations can
be defined among these classes using the XML attribute RefBaseClassPath. The
next step is the modeling of system units which represent reusable engineering
objects, for example the robot model kr5 from the manufacturer KUKA. Each
system unit may consist of interfaces (called as external interfaces) and nested in-
ternal structures (called as internal elements). The composition is illustrated by
the connections with a filled diamond endpoint in figure 1. The star symbol over
Role InterfaceClass
+ Name: String + Name: String
+ RefBaseClassPath: String + RefBaseClassPath: String
references references
* SystemUnit hasInterface ExternalInterface
*
+ Name: String + Name: String
hasAttribute hasAttribute
hasInternalElement * *
Attribute
+ Name: String
Fig. 1. The XML scheme of AutomationML. Some details are omitted for brevity.
one connection means that the cardinality of the composition is unlimited. Op-
tionally, the semantics of a system unit or an external interface can be specified
as a reference to an AML role or interface class respectively. While an internal el-
ement and an external interface can reference only one single class, a system unit
can reference more than one to allow the modeling of multi-functional devices.
Finally, attributes can be added to describe properties of objects, e.g. kr5.weight.
In order to learn concepts from AML, we firstly need a formal semantic
representation of the XML-based data. The Web Ontology Language (OWL) is
the W3C standard for knowledge representation on the web and its subset OWL
2 DL is semantically compatible with the SROIQ description logic. Besides
of its decidability, OWL 2 DL provides a rich vocabulary for complex class
expressions, making it a rational choice for the semantic lifting. In particular,
the support of nominals and concrete roles allows us to describe meaningful
engineering concepts. For example, robots which have at least 6 axis from the
manufacturer KUKA can be stated as:
Robot ⊓ ∃hasManufacturer.[”KUKA”] ⊓ ∃hasNumAxis.[≥ 6] (1)
Table 1 shows the mapping strategy from AML to OWL 2 DL. Specifically,
AML role and interface classes are mapped to OWL classes, while system units
and external interfaces are mapped to OWL individuals. Intuitively, relations
between objects are mapped to object properties, and attributes are mapped to
data properties. In this paper, we restrict the scope of the lifted AML ontol-
ogy and focus on just two object properties: hasInternalElement (hasIE in short)
and hasExternalInterface (hasEI in short). Because internal elements and exter-
nal interfaces can be independent of any AML class, both hasIE and hasEI have
the range as OWL : Thing. The abstract conceptual model of AML is not trans-
formed since it does not carry useful semantics for engineering. However, we add
an annotation to each lifted AML entity to indicate its role in the XML schema.
For example, each lifted system unit has an annotation of SystemUnit. After the
transformation, we obtain a knowledge base K = (T , A) including a terminolog-
ical part T (T-Box) and an assertional part A (A-Box). The T-Box contains the
concept definitions of AML role and interface classes, and the A-Box contains
AML OWL 2 DL DL Example
role class class concept Robot
interface class class concept SignalInterface
system unit individual object kr5
external interface individual object kr5 digitalIn1
relationship object property abstract role hasIE, hasEI
attribute data property concrete role hasWeight
Table 1. Overview of the conceptual mapping from AML to OWL 2 DL.
the ground facts, i.e. system units and their internal structures. For example,
following assertions are generated for the system unit kr5:
hasManufacturer(kr5, ”KUKA”)
hasExternalInterface(kr5, kr5 digitalIn1)
hasInternalElement(kr5, kr5 arm)
DigitalIOInterface(kr5 digitalIn1)
Robot(kr5 arm) (2)
3 Concept learning in AutomationML
Concept Learning is a subfield of machine learning which aims to induce a con-
cept description for a set of positive and negative examples. In this paper, we
focus on concept learning in description logics, and follow the definition of a
learning problem as proposed in [10].
Definition 1 (concept learning in description logics). Let Target be a con-
cept name and K be a knowledge base (not containing Target). Let E = E + ∪ E −
be a set of examples, where E + are the positives examples and E − are the
negative examples. The learning problem is to find a concept C ≡ Target with
K ∪ C |= E + and K ∪ C 6|= E − .
In the context of AML, the ultimate goal is to learn the description of a named
class from a set of AML system units, since they are models of engineering objects
that we are interested in. Figure 2 illustrates the learning procedure. While the
AML data is mapped to an OWL 2 DL ontology as described before, the user
is expected to select some AML system units as positive examples E + , while
leaving the rest as the negatives E − . Afterwards, a configuration file has to be
generated for the learning system.
Concept Learning methods often employ refinement operators to reduce the
search space of concept hypotheses. The essential property of the refinement
operator ρ in DL-Learner is its completeness in ALC. That means, starting from
any concept C (including ⊤), ρ will reach the target concept with sufficient
time. Although the support of concrete roles makes ρ incomplete because of
Examples
select pos pos pos
AML Config
neg neg
map to setting
Background Knowledge
OWL DL-Learner
Fig. 2. Procedure of concept learning in AutomationML.
the infiniteness of real numbers, ρ is capable of finding proper concrete roles by
computing splits in the space of real numbers. However, the direct use of ρ would
be inefficient, because it does not take into account the syntactic constraints
defined in the underlying XML schema of AML. Specifically, figure 1 shows that
only external interfaces can reference interface classes and each external interface
can only reference one interface class. These constraints can be integrated into
the refinement operator to improve the performance of learning, especially for a
large T-Box. Therefore, we divide the set of named concepts NC into two subsets
Nar and Nai , where Nar is the set of all AML role classes and Nai is the set of
all AML interface classes. Then we define the sets Mop , Mie and Mei as follows:
Mop = {∃hasIE.⊤, ∃hasEI.⊤, ∀hasIE.⊤, ∀hasEI.⊤} (3)
′ ′
Mie = {A | A ∈ Nar , ∄A ∈ Nar : A ⊏ A } (4)
Mei = {A | A ∈ Nai , ∄A′ ∈ Nai : A ⊏ A′ } (5)
Mie is the set of top level AML role classes, and Mei is the set of top level
AML interface classes. Further, let Uie = {C1 ⊔ C2 ⊔ ... | Ci ∈ Mie ∪ Mop },
Uei = {C1 ⊔ C2 ⊔ ... | Ci ∈ Mei } and sh↓ (C) be the set of direct sub classes of a
named concept C ∈ NC , we extend ρ in the following cases:
– ρ(C) = Uei , if C = ⊤ and C is a filler of hasEI
– ρ(C) = Uie , if C = ⊤ and C is not a filler of hasEI
– ρ(C) = sh↓ (C), if C ∈ NC is a filler of hasEI
– ρ(C) = sh↓ (C) ∪ {C ⊓ D|D ∈ ρ(⊤)}, if C ∈ NC is not a filler of hasEI
In the other cases, we keep ρ as it was in DL-Learner and omit the details
of it in this paper for brevity. We call the new refinement operator ρaml and
implemented it based on ρ in the DL-Learner framework. Notice that negated
atomic concepts such as ¬A are ignored in both Mie and Mei , since negations are
not preferred in engineering and are not used in practice. Moreover, we do not
refine the filler of hasEI with concept intersections, because one external interface
can reference only one interface class and does not have nested structures. As a
result, ρaml would generate far less refinements than ρ. For example, a refinement
chain ⊤ ∃hasEI.⊤ ∃hasEI.C1 ∃hasEI.(C1 ⊓ C2 ) could be produced by ρ
but not by ρaml , since (C1 ⊓ C2 ) is not a proper refinement of the filler of hasEI.
Benchmark 1 Benchmark 2
Concepts Overall Reasoning Concepts Overall Reasoning
T1 3996(65%) 508(75%) 304(73%) 4733(54%) 451(64%) 308(68%)
T3 4926(13%) 595(25%) 404(28%) 5746(10%) 800(15%) 541(18%)
T17 67614(5%) 5483(5%) 3785(6%) 113914(6%) 9243(5%) 6361(6%)
T12 67918(4%) 5465(4%) 3741(5%) 114427(7%) 8792(8%) 6170(9%)
T14 562705(?%) 84419(?%) 40997(?%) 812225(?%) 105779(?%) 53679(?%)
Table 2. Evaluation of CELOE using both refinement operators.
4 Results
In this section, we compare learning results using the refinement operator ρ
from DL-Learner and the extended one ρaml as described above. Specifically, we
measure the time required until the first 100% accurate solution is found. The
raw AML document comes from the research project ReApp which was origi-
nally used for modeling industrial robot systems [7]. The lifted AML ontology
comprises of 222 classes, 497 individuals and 73 data properties. To simulate the
heterogeneity in engineering projects, we perform two different benchmarks. The
first one has an additional 50 AML role and 25 AML interface classes, while the
second one has twice as much. We choose the algorithm CELOE, since its heuris-
tic is configured to produce shorter concepts than OCEL. Generally speaking,
CELOE sets a stronger penalty to long refinements and is less rewarded by the
immediate accuracy gain. In each benchmark, we run 25 experiments of different
target concepts. Axiom 6 show one example of the ground truth.
RobotWithServoMotors ≡ ∃hasIE.(ArticulatedRobot ⊓ ∃hasIE.ServoMotor) (6)
Table 2 summarizes the results of five representative experiments of varying
complexity, while the rest ones share similar characteristics and are omitted for
brevity. The column Concepts is the number of tested concept hypotheses. The
column Overall is the overall time needed to find the first correct solution, and
the column Reasoning is the time spent for reasoning, both timers are in millisec-
onds. All measurements in the table are taken from ρaml , while the percentages
in parentheses represent the ratio in the form of ρaml /ρ. Unknown values of the
ratio means that no solution was found within 10 minutes with ρ. The results
show that ρaml is significant faster than ρ (the ratios are much smaller than
100%), since ρaml generates much less concept hypotheses for testing. However,
for some cases no solution was found within 10 minutes with either ρ or ρaml ,
so learning in AML still remains a challenging task2 .
5 Conclusion
In this paper, we studied concept learning from engineering data stored in the
XML-based data format AML. We showed how to use the DL-Learner framework
2
By reducing the expansion penalty of CELOE, we were able to find a solution with
the extended refinement operator ρaml for some of these hard cases.
to learn engineering concepts in description logic, by transforming an AML docu-
ment to an OWL 2 DL ontology. We proposed an extension of the ALC downward
refinement operator ρ in DL-Learner to exploit the syntactic constraints defined
in the XML schema of AML. Experimental results show that the extended oper-
ator ρaml has a significant performance improvement in all test cases. However,
learning is still very challenging in some cases and could be worse if the size
of the T-Box grows further. In future work, we want to investigate whether we
can achieve better learning results using a semantic language other than OWL
2 DL, for example the hybrid system AL-log that merges ALC and DATALOG.
In particular, an ideal refinement operator for AL-log was proposed in [11]. For
learning in OWL 2 DL, we want to study if a heuristic can intelligently adapt
itself to avoid the laborious fine tuning of learning parameters.
References
1. Abele, L., Legat, C., Grimm, S., Müller, A.W.: Ontology-based validation of plant
models. In: 2013 11th IEEE International Conference on Industrial Informatics
(INDIN). pp. 236–241 (July 2013)
2. Bigvand, P.G., Drath, R., Scholz, A., Schüller, A.: Agile standardization by means
of PCE requests. In: 2015 IEEE 20th Conference on Emerging Technologies Factory
Automation (ETFA). pp. 1–8 (Sept 2015)
3. Bigvand, P.G., Fay, A., Drath, R., Carrion, P.R.: Concept and development of a
semantic based data hub between process design and automation system engineer-
ing tools. In: 2016 IEEE 21st International Conference on Emerging Technologies
and Factory Automation (ETFA). pp. 1–8 (Sept 2016)
4. Bühmann, L., Lehmann, J., Westphal, P.: DL-Learner: A framework for inductive
learning on the semantic web. Web Semant. 39(C), 15–24 (Aug 2016)
5. Drath, R., Lüder, A., Peschke, J., Hundt, L.: AutomationML - the glue for seam-
less automation engineering. In: 2008 IEEE International Conference on Emerging
Technologies and Factory Automation. pp. 616–623 (Sept 2008)
6. Hua, Y., Hein, B.: Concept learning in AutomationML with formal semantics and
inductive logic programming (accepted). In: 14th IEEE International Conference
on Automation Science and Engineering (Aug 2018)
7. Hua, Y., Zander, S., Bordignon, M., Hein, B.: From AutomationML to ROS: A
model-driven approach for software engineering of industrial robotics using onto-
logical reasoning. In: 2016 IEEE 21st International Conference on Emerging Tech-
nologies and Factory Automation (ETFA). pp. 1–8 (Sept 2016)
8. Kovalenko, O., Wimmer, M., Sabou, M., Lüder, A., Ekaputra, F., Biffl, S.: Mod-
eling AutomationML: semantic web technologies vs. model-driven engineering. In:
Emerging Technologies Factory Automation (ETFA), 2015 IEEE 20th Conference
on. pp. 1–4 (Sept 2015)
9. Lehmann, J., Auer, S., Bühmann, L., Tramp, S.: Class expression learning for
ontology engineering. Web Semantics: Science, Services and Agents on the World
Wide Web 9(1), 71 – 81 (2011)
10. Lehmann, J., Hitzler, P.: Concept learning in description logics using refinement
operators. Machine Learning 78(1), 203 (Sep 2009)
11. Lisi, F.A., Malerba, D.: Ideal refinement of descriptions in AL-log. In: Horváth, T.,
Yamamoto, A. (eds.) Inductive Logic Programming. pp. 215–232. Springer Berlin
Heidelberg, Berlin, Heidelberg (2003)