GECCO: GERMAN CRAFTS & CRAFTSMEN ONTOLOGY

A Common Crafts Ontology

Andrea Horch and Maximilien Kintz

Institute of Human Factors and Technology Management, IAT University of Stuttgart

Nobelstr. 12, 70569 Stuttgart, Germany

Keywords:

Ontology, Semantic Web, Crafts, Craftsmen.

Abstract:

In this paper we introduce GeCCO (German Crafts & Craftsmen Ontology), an ontology to establish a common

crafts classiﬁcation, to be used for example in craftsmen search systems. Current standardized categorization

systems, which order crafts, do not include all needed logical connections between crafts and craftsmen as

well as the synonyms for a sufﬁcient use in a search system. Actual craftsmen search systems have their own

categorizations of crafts and craftsmen, but they have the same problems as the standardized systems.

The crafts ontology proposed in this paper solves the mentioned problems by creating an extensible class

hierarchy as well as by deﬁning synonyms and all needed logical connections between the classes. For quality

assurance the ontology was evaluated by crafts experts.

1 INTRODUCTION

There are several search platforms and online market-

places like MyHammer

or blauarbeit.de

for search-

ing craftsmen. Most of them are organized by key-

word search combined with a search by various cate-

gories. These platforms are not using a consistent and

standardized form to categorize crafts.

Currently in Germany there are two popular cate-

gorization systems ordering crafts:

1. The German Crafts Code

(Handwerksordnung),

which lists more than 100 crafts and crafts similar

trades.

2. The eCl@ss

classiﬁcation system for products

and services.

These systems are not sufﬁcient for use in crafts

search services since they do not include enough log-

ical connections between a craft and the associated

craftsmen. The German Crafts Code does not de-

scribe the overlap of craftsmen having common activ-

ities. eCl@ss shows only some connections between

a craft and its performing craftsmen in the form of

keywords.

http://news.myhammer.com/

http://www.blauarbeit.de/

http://www.gesetze-im-internet.de/hwo/

http://www.eclass.de/

Thus there are no existing crafts categorization sys-

tems sufﬁcient for the use in search systems. The on-

tology proposed in this paper shall close this gap. We

named it GeCCO (German Crafts & Craftsmen On-

tology) since it is modeled on the structure of the ger-

man crafts like given in the German Crafts Code.

The paper is structured as follows: section 2 ex-

plains the limitations of current crafts classiﬁcations

and search systems. Section 3 gives some examples

for related work. Section 4 introduces the approach

for the ontology development. The proposed concepts

of the ontology are presented in section 5. The eval-

uation is shown in section 6. We conclude and give

some future prospects in section 7.

2 MOTIVATION

The common crafts ontology proposed in this pa-

per focuses on two aspects: the creation of a con-

sistent crafts categorization and the development of

a crafts model including all logical connections be-

tween crafts and the associated craftsmen by using

semantic technologies.

Testing ﬁve of the most popular online craftsmen

search systems (see table 1) we identiﬁed the follow-

ing current problems in their crafts categorizations:

1. Redundancy of Classes: a craftsman class has

more than one master category. There is only

355

Horch A. and Kintz M..

GECCO: GERMAN CRAFTS & CRAFTSMEN ONTOLOGY - A Common Crafts Ontology.

DOI: 10.5220/0003902003550360

In Proceedings of the 8th International Conference on Web Information Systems and Technologies (WEBIST-2012), pages 355-360

ISBN: 978-989-8565-08-2

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

a hierarchical connection between the craftsman

class (presented in each master class) and its di-

rect master class, but there are no logical con-

nections between the craftsman class and all of

its master classes (see ﬁgure 1) or between these

craftsman classes among each other.

Glasswork

Windows & Doors

Joinery & Carpentry

Glazier

Joiner / Carpenter

…

Joiner / Carpenter

…

Figure 1: Redundancy of craftsman classes.

This kind of classiﬁcation returns different search

results for each instance of the craftsman class.

2. Fuzzy Classiﬁcation: a craftsman class is not as-

signed to the expected master class. Instead, it is

categorized in a fuzzy named class like "Miscel-

laneous" or "Other" (see ﬁgure 2).

Construction Work

Miscellaneous

Scaffolder

Bricklayer

Plasterer

Cycle Mechanic

Roofer

…

Figure 2: Fuzzy classiﬁcation.

Thus, the craftsman class is very hard to ﬁnd.

3. Disregard for Equivalent Classes: equivalent

craftsman classes are assigned to different master

classes and there is no logical connection between

them (see ﬁgure 3).

Glasswork

Windows & Doors

Glazier

Window Builder

…

Joiner / Carpenter

…

Figure 3: Disregard for equivalent classes.

That means the equivalent classes are completely

separated from each other therefore they return

different search results.

Other problems are missing synonyms and unre-

garded changes in the structure of a craft like the

consolidation of some craft trades to one trade (in

Germany the trades "forger" and "locksmith" were

combined to the trade "metal worker" in the 1980s

(Berufe-Lexikon, 2011)).

For solving all introduced problems a system rep-

resenting the objects of the crafts domain (classiﬁca-

tion) and the relationships among them is needed. For

Table 1: Tested crafts search systems.

Platform Name URL

MyHammer http://www.myhammer.com

Find a Craftsman http://www.ﬁndacraftsman.com

blauarbeit.de http://www.blauarbeit.de

handwerker24 http://www.handwerker24.de

jobdoo http://www.jobdoo.de

this purpose an ontology was created and is proposed

in this paper. An ontology is deﬁned as "an explicit

speciﬁcation of a conceptualization" (Gruber, 1993),

which represents the domain knowledge including all

objects and their relationships.

3 RELATED WORK

The ASTECH project (Pralon and Million-Rousseau,

2011) deﬁnes a craft ontology and terminologies for

knowledge sharing in the sector of the renewable en-

ergies area across Europe. The Web Ontology for

Products and Services (eClassOWL

) the semantic

model of the eCl@ss classifcation system (see section

1) is an approach to classify products and services,

which also includes some crafts services.

4 DESIGN APPROACH

(Uschold and King, 1995) introduce a methodology

that presents several recommendations and steps for

building ontologies. Some analysis of this method-

ology done in (López, 1999) results in the ﬁnding,

that the methodology does not propose a life cycle for

ontology design, which could be a problem when ex-

tending or reengineering the ontology for purposes of

quality assurance.

Nevertheless we have chosen the mentioned

methodology for creating our ontology since it is a

purpose driven approach of ontology design. In some

steps we adapted it a bit and integrated a cycle for

the purposes of quality assurance. After including our

adjustments the methodology results in the following

steps for ontology design (see ﬁgure 4):

1. Identiﬁcation of Purpose(s).

(Uschold and King, 1995) determine that clarify-

ing the purpose(s) of the ontology (use cases and

users) is important for the ontology design.

2. Ontology Capture.

This step covers the identiﬁcation of the key con-

cepts like classes, class hierarchies and the rela-

http://www.heppnetz.de/projects/eclassowl/

WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies

356

1. Identification

of purpose(s)

2. Ontology

Capture

3. Building the

Ontology Concept

4. Evaluation by

Domain Experts

5. Coding /

Modelling of the

Ontology

6. Evaluation of

Consistency

Figure 4: Ontology design steps.

tionships among them. Additionally it includes

the deﬁnition of clear textual names for the classes

and relations (Uschold and King, 1995).

3. Building the Ontology Concept.

This part of ontology design contains the creation

of triples (subject, predicate, object; for more in-

formation see (Segaran et al., 2009)) to sketch the

ontological concept (classes, class hierarchies, re-

lationships) in a form readable and understand-

able for domain experts (who usually are no on-

tology experts); like e.g. tables of an Excel sheet.

4. Evaluation by Domain Experts.

Some domain experts evaluate the created con-

cept. If necessary, they do some corrections and

provide suggestions for improvement. This step

and step No. 3 can be repeated till the experts

conﬁrm the correctness of the ontology concept.

5. Coding/Modeling of the Ontology.

After the ﬁrst evaluation step the ontology be-

comes coded or modeled in an ontology descrip-

tion language like OWL (W3C, 2009) or DAML

+ OIL (Stevens et al., 2003) including all logical

relationships.

6. Evaluation of Consistency.

Check the ontology for conceptual consistency;

e.g. by using a reasoner like HermiT (Shearer

et al., 2008) or Pellet (Sirin and Parsia, 2004).

5 CRAFTS ONTOLOGY

To model the ontology we decided to use Protégé

since it is an open source editor, which underlies a

continuous development process. Furthermore Pro-

tégé offers support as well as a lot of features like

several reasoners (Horridge et al., 2007).

The ontology was built in the standardized Web

Ontology Language (OWL 2 DL) (W3C, 2009).

Some parts of the ontology we want to describe in

detail in this paper are held in Manchester OWL Syn-

tax (W3C, 2009) (Horridge et al., 2006) and some de-

scription logics (Baader et al., 2008) for clarity.

The basis of the ontology is built by the four main

classes Craft, Craftsman, Object and TypeOfLoss (see

ﬁgure 5), which are disjoint from each other.

Figure 5: Ontology structure.

These four main classes have all the same structure.

They are subclasses of the class Thing and have an

English and a German label.

Class : Cr aft

Ann otations :

l abel " c r af t "@en

l abel " Handwerk "@de,

SubClassOf :

Thing

DisjointWith :

Craftsman , Object , TypeOfLoss

Listing 1: Class craft.

GECCO:GERMANCRAFTS&CRAFTSMENONTOLOGY-ACommonCraftsOntology

357

The class Craft (see listing 1) is the main class of all

different crafts like masonry, rooﬁng work, etc. The

class Craftsman is the main class of all craftsmen like

a mason or a roofer. The class Object comprises the

classes of objects, which can be processed by a crafts-

man. TypeOfLoss is the main class of the different

types of losses, which can damage an object; e.g. the

breakage of glass or a ﬁre loss.

Like shown in ﬁgure 5 the classes of the ontol-

ogy are restricted by eight different object properties.

Each two of the deﬁned object properties are inverse.

The object property processes and its inverse

isProcessedBy deﬁne which craftsman processes

which object(s) (see listing 2).

ObjectProperty : pro ces s es

Ann otations :

l ab e l " p roc es s es "@en,

l ab e l " b e a r b e i t e t "@de

C h a r a c t e ri s t i cs :

Asymmetric ,

I r r e f l e x i v e

Domain :

Craftsman

Range :

Object

InverseOf : ’ i s process ed by ’

Listing 2: Object property processes.

ObjectProperty : ’ i s component of ’

Ann otations :

l ab e l " i s component of "@en,

l ab e l " i s t B e s t an d teil von "@de

C h a r a c t e ri s t i cs :

T r an s i ti v e

Domain :

Object

Range :

Object

InverseOf : ’ c o n si s t s of ’

Listing 3: Object property isComponentOf.

The object property offersService and its inverse

isOfferedBy describe which craftsman offers which

craft.

Which type of loss often damages which object(s)

is given by the object property damagesOften and its

inverse oftenDamagedBy.

The structure of the object properties offersSer-

vice, damagesOften and their inverses is equal to the

structure of the object property processes and its in-

verse. The six depicted object properties are all de-

ﬁned as asymmetric and irreﬂexive.

The object property isComponentOf (see listing

3) and its inverse consistsOf is a special property,

because it deﬁnes which object consists of which

other object(s). That means, the object property is

transitive; so domain as well as range refer to the

class Object.

All crafts are deﬁned as subclasses of the class

Craft. The class Masonry, for instance, describes the

characteristics of the craft masonry (see equation 1).

Looking at the deﬁnition of the class Masonry

one can see that masonry is a craft offered by a ma-

son or a concrete worker (object property: ’is service

of’). The keyword "some" deﬁnes that it exists at

least one mason (class Mason) and at least some con-

crete worker (class ConcreteWorker) whose offered

services include masonry.

Masonry ≡ Craft ⊓ ∃isServiceOf.Mason

⊓ ∃isServiceOf.ConcreteWorker

(1)

It is impossible to claim that there are no other crafts-

men offering masonry work. Craft is very complex,

it is imprecisely deﬁned which craftsmen offer which

crafts and some crafts overlap in some activities.

Thus, class restrictions can only be deﬁned by the

quantiﬁer ∃, which deﬁnes existential restrictions,

and may not be deﬁned by ∀ deﬁning universal

restrictions.

The subclasses of the main class Craftsman de-

ﬁne the different craftsmen, e.g. the class Mason.

The logical deﬁnition of the class Mason is given by

the following term:

Mason ≡ ConcreteWorker ≡ (Craftsman

⊓ ∃of fersService.Masonry

⊓ ∃of fersService.NaturalStoneProcessing

⊓ ∃of fersService.InsulationWork

⊓ ∃of fersService.DiaphragmWallWork

⊓ ∃of fersService.ScreedWork)

(2)

StormLoss is an example for a subclass of the main

class TypeOfLoss. These subclasses deﬁne the types

of loss, which often damage some objects (equation

3). For determining the objects a type of loss often

damages, one also has to use an existential restriction,

because a type of loss can damage much more objects

than can be deﬁned by the ontology.

WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies

358

From the subclasses of the main class Object one

can get lots of information like which types of loss

often damages the object, which craftsmen process it

and if it is a component of another object or if it con-

sists of other objects. An exampleis given by the class

WindowGlass (see equation 4).

StormLoss ≡ (TypeOfLoss

⊓ ∃damagesOften.Fence

⊓ ∃damagesOften.Roof

⊓ ∃damagesOften.Window

⊓ ∃damagesOften.RoofCovering

⊓ ∃damagesOften.WindowGlass)

(3)

WindowGlass ≡ (BuildingElement

⊓ ∃isProcessedBy.Glazier

⊓ ∃isProcessedBy.WindowFitter

⊓ ∃isComponentOf.Window

⊓ ∃oftenDamagedBy.BreakageOfGlass

⊓ ∃oftenDamagedBy.StormLoss

⊓ ∃oftenDamagedBy.Vandalism

⊓ ∃oftenDamagedBy.Burglary

⊓ ∃oftenDamagedBy.Earthquake

⊓ ∃oftenDamagedBy.Landslide

⊓ ∃oftenDamagedBy.FireLoss

⊓ ∃oftenDamagedBy. Avalanche)

(4)

The object property isComponentOf and its inverse

consistsOf are transitive. For example: a tile consists

of clay and a tile roof consists of tiles, which implies

that a tile roof also consists of clay.

SolarEnergySpecialist ≡ (Craftsman

⊓ (AirHeatingFitter ⊔ Electrician

⊔ ElectronicsTechnician⊔ HeatingEngineer

⊔ Roofer⊔ StoveEngineer)

⊓ ∃of ferService.SolarEngineering

⊓ ∃processes.SolarHotWaterSystem

⊓ ∃processes.PhotovoltaicSystem)

(5)

GeCCO models synonyms as labels (classes have dif-

ferent English and German labels to describe syn-

onyms). Examples for synonyms are the terms joiner

and cabinet maker or masonry and walling. Special

craftsmen are subclasses of other craftsman classes;

e.g. solar expert is subclass of roofer or electrician

(equation 5).

GeCCO deﬁnes a total of 524 classes and 8 ob-

ject properties. The main class Craft has 106 sub-

classes. The main class Craftsman has 185 sub-

classes. 211 classes are subclasses of the main

class Object, whereas the last main class TypeOfLoss

counts 17 subclasses.

6 EVALUATION

From the design criteria described in section 4 and

the ones mentioned in (Gruber, 1993) we can derive

the following criteria for doing the evaluation:

1. Structural/Technical Consistency.

This ﬁrst point affects the technical correctness of

the ontology. That means, if it is structurally as

well as logically correct. GeCCO was tested for

its consistency by using three different reasoners:

HermiT

, Pellet

and FaCT++

. These evaluation

steps meet the criteria of Clarity, Coherence and

Extendibility described in (Gruber, 1993) and the

Ontology Capture step of section 4.

2. Functional/Textual Consistency.

Functional / textual consistency means the accu-

racy of the ontology content, or rather the ontol-

ogy concept. That affects the ontology class hier-

archy, the logical connections between the classes

as well as the design of synonyms.

For these purposes we created an Excel sheet in-

cluding the crafts and craftsman classes of the on-

tology as well as the connections between them

in form of different tables. We designed it in the

form of triples (subject, predicate, object). The

ﬁrst design was made in Excel since it was the

format of choice of our domain experts. These

domain experts, who supported the creation of the

ontology domain concept, were employees of the

Chamber of Crafts of Stuttgart. Ontology model-

ing activities in OWL did not start before the eval-

uation of the Excel version of the domain concept

was ﬁnished.

This evaluation approach ﬁts the design criteria of

Minimal encoding bias and Minimal ontological

commitment as introduced in (Gruber, 1993) and

the Ontology Concept step of section 4.

3. Practical Suitability.

In the ﬁrst design step of section 4 we mentioned

the proposition of (Uschold and King, 1995) , who

determine that it is essential for ontology design

to clarify the purpose(s) of the ontology. From

this determination we derive the importance for

ontology evaluation to examine the suitability of

the created ontology for the purpose of its creation

deﬁned in the ﬁrst step of the ontology design.

Therefore we developed a semantic craftsmen

search based on GeCCO, which we tested for

different search queries. The craftsmen search

http://hermit-reasoner.com/

http://clarkparsia.com/pellet/

http://owl.man.ac.uk/factplusplus/

GECCO:GERMANCRAFTS&CRAFTSMENONTOLOGY-ACommonCraftsOntology

359

is connected to a test-database containing real

craftsmen data.

The main problem GeCCO focuses on is the fact

that laymen often do not know which craftsman is

processing which damage. GeCCO helps a user of

a craftsmen search system to ﬁnd the right crafts-

man for eliminating a damage. Therefore a user

can just type in the name of the damaged item

or the craft he is searching for as a keyword into

the semantic search and GeCCO will ﬁnd the right

type(s) of craftsmen for doing the database search.

The evaluation of GeCCO has shown that GeCCO is

structural and technical as well as functional and tex-

tual consistent. The test for the practical suitability

for the use case of a craftsmen search system demon-

strated the comfort of a semantic system based on a

domain ontology.

7 CONCLUSIONS/FUTURE

WORK

We have introduced GeCCO, an ontology describing

crafts. GeCCO delivers a good class basis for crafts

search services or crafts translation services, but does

not claim to be complete.

In this paper we have also shown how to evalu-

ate the ontology. We have checked the ontology for

structural consistency and evaluated the concept by

the support of domain experts. The ontology was

tested as basis of a semantic craftsmen search and has

passed the test. The test has shown the comfort of an

ontology-based user guidance for web searches, espe-

cially for complex search domains like crafts.

Further work should focus on the maintanance and

the extension of GeCCO. One way to extend the on-

tology could be the use of GeCCO in an online search

system in conjunction with an automatism, which

learns from user entries and extends GeCCO automat-

ically by using the acquired knowledge.

Furthermore GeCCO could be extented by other

modules like eClassOWL (see section 3), e.g. by em-

bedding the product-classes of eClassOWL as object-

classes into GeCCO.

ACKNOWLEDGEMENTS

The work published in this article was partially

funded by the openXchange project of the German

Federal Ministry of Economy and Technology under

the promotional reference 01MQ09011.

REFERENCES

Baader, F., Horrocks, I., and Sattler, U. (2008). Description

Logics. In van Harmelen, F., Lifschitz, V., and Porter,

B., editors, Handbook of Knowledge Representation,

chapter 3, pages 135–179. Elsevier B.V.

Berufe-Lexikon (2011). Berufsbild Schlosser/in

@http://www.berufe-lexikon.de/berufsbild-

beruf-schlosser.htm. http://www.berufe-

lexikon.de/berufsbild-beruf-schlosser.htm.

Gruber, T. R. (March 1993). Toward principles for the de-

sign of ontologies used for knowledge sharing. In In

Formal Ontology in Conceptual Analysis and Knowl-

edge Representation. Kluwer Academic. Available

as Stanford Knowledge Systems Laboratory Report

KSL-93-04.

Horridge, M., Drummond, N., Goodwin, J., Rector, A.,

Stevens, R., and Wang, H. H. (2006). The manchester

owl syntax. Technical report, WebOnt.org - A WEB

ONTOLOGY PORTAL, The University of Manch-

ester.

Horridge, M., Jupp, S., Moulton, G., Rector, A., Stevens,

R., and Wroe, C. (2007). A practical guide to building

owl ontologies using protégé 4 and co-ode tools.

López, M. F. (1999). Overview of Methodologies for Build-

ing Ontologies. In Proceedings of the IJCAI-99 Work-

shop on Ontologies and Problem Solving Methods

(KRR5) Stockholm, Sweden, August 2, 1999.

Pralon, S. and Million-Rousseau, C. (2011). Astech project:

craft ontology and terminologies to share knowledge

in renewable energies area across europe. Interna-

tional Journal of Reasoning-based Intelligent Systems

(IJRIS), Volume 3 - Issue 3/4 - 2011.

Segaran, T., Evans, C., and Taylor, J. (2009). Programming

the Semantic Web. O’Reilly, Wiesbaden.

Shearer, R., Motik, B., and Horrocks, I. (2008). Hermit: A

highly-efﬁcient owl reasoner. In Dolbear, C., Rutten-

berg, A., and Sattler, U., editors, OWLED, volume 432

of CEUR Workshop Proceedings. CEUR-WS.org.

Sirin, E. and Parsia, B. (2004). Pellet: An owl dl reasoner.

In Description Logics.

Stevens, R., Wroe, C., Bechhofer, S., Rector, A., and Goble,

C. (2003). Building Ontologies in DAML+OIL. Com-

parative and Functional Genomics, 4(1):133–141.

Uschold, M. and King, M. (1995). Towards a methodology

for building ontologies. In Workshop on Basic Onto-

logical Issues in Knowledge Sharing, held in conjunc-

tion with IJCAI-95, Montreal, Canada.

W3C (2009). OWL 2 Web Ontology Language Document

Overview.

W3C (2009). OWL 2 web ontology language manchester

syntax. http://www.w3.org/TR/owl2-manchester-

syntax/.

WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies

360