Development and Population of an Elaborate Formal Ontology for

Clinical Practice Knowledge Representation

David Mendes

, Irene Pimenta Rodrigues

and Carlos Fernandes Baeta

Departamento de Informática, Universidade de Évora, 7004-516 Évora, Portugal

Departmento de Medicina, Hospital José Maria Grande, Portalegre, Portugal

Keywords:

OGCP, OGMS, CPR, Ontological Realism, SOAP, Clinical Practice Knowledge, OWL.

Abstract:

We introduce the Ontology for General Clinical Practice (OGCP) for better knowledge representation support

in the Clinical Practice domain. We followed the established OBO Foundry principles to leverage the onto-

logical relations that might be present in the ontology axioms we harvest from clinical reports text segments.

In accordance to the Ontological Realism principles we expect the reasoning inferred from the ontological

relations to render more acceptable consequences then logical relations alone. We enhance the Ontology for

General Medical Science (OGMS) with the Computer-Based Patient Record Ontology (CPR) structure and

propose knowledge base creation/enhancing automatically extracting from clinical reports written in the, well

known to the medical community, SOAP format. Reasoning over the resulting (OGCP) knowledge base with

novel parallel algorithms that appeared recently in literature is presented. We ﬁnally propose Controlled

Natural Language justiﬁcations of the inferred knowledge intending to achieve wider acceptance among clin-

icians.

1 MOTIVATION AND RESEARCH

QUESTIONS

Originally our research intention was the develop-

ment of personal CDS

tools to help the healthcare

professionals in scarce resource countries like most

in Africa and Asia. After evaluating the State-of-the-

Art presented ahead we found that relevant work is

yet to be done in the KR

area regarding the Clinical

Practice domain. We believe that some developments

that have been achieved recently motivate us to incor-

porate our expertise in NLP

into effective ontology

population. Our main intention is to be able to auto-

matically produce clinical practice knowledge bases

extracting from healthcare reports text.

Research Questions. Ontologies in the sub-domain

of Clinical Medicine

are lacking some thorough

study. These can be stated as current problems for the

effectiveness of using them as knowledge support for

Clinical Decision Support

Knowledge Representation

Natural Language Processing

The study of disease by direct examination of the living

patient

clinical reasoning. Problems found in current ontolo-

gies and enumerated in literature (Hoehndorf et al.,

2011) that lead to reasoning hurdles are:

• Lack of adequate modularization to achieve

the minimum amount of implicit differentiation

among primitive concepts.

• Inadequate clear separation of digital entities

from the reality they represent.

• Inability to avoid the knowledge acquisition bot-

tleneck (Wong et al., 2012) in order to speed start

any automatic enrichment.

In our work we try to overcome the different is-

sues identiﬁed by the several experts in (Brochhausen

et al., 2011). In order to maximize the reasoning capa-

bilities based in our extended OGMS(OGMS, 2010)

ontology, different considerations in the referred work

by Brochhausen et al. were taken into good ac-

count. We complemented the OGMS ontology with

the CPR into what we call the OGCP that is intended

to be a more supportive structure for representation of

clinical practice while, at the same time, embodies a

formal medical theory of disease and healthcare.

286

Mendes D., Rodrigues I. and Fernandes Baeta C..

Development and Population of an Elaborate Formal Ontology for Clinical Practice Knowledge Representation.

DOI: 10.5220/0004548602860292

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2013), pages 286-292

ISBN: 978-989-8565-81-5

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

2 STATE OF THE ART

Our previous work (Mendes and Rodrigues, 2012b),

where an exhaustive state of the art is presented, is

focusing mainly over the CPR ontology (W3C, 2009)

but we are now targeting an extended OGMS because

it is more promising as suitable for representation

of a disease theory and model enhancing the corre-

sponding reasoning capabilities. We base our work

in (Ogbuji, 2011) for what matters about the foun-

dational principles of structuring meaningful knowl-

edge representation as a framework for clinical rea-

soning. We consider the above mentioned (Smith and

Ceusters, 2010) for Ontological Realism approach.

We used the excellent recent review by Wong (Wong

et al., 2012) for updated state of the art problems in

knowledge acquisition from text. We explore the re-

cent achievements in controlled natural language gen-

eration presented in (Kaljurand, 2010) with the dis-

tributed processing possibilities suggested in (Kaza-

kov et al., 2011) for consequence based axiom infer-

ring introduced in (Simancik et al., 2011).

3 APPROACH

We illustrate the relations between open questions

and the current line of work to illustrate the used ap-

proach:

Ontological Realism and Relations. The method-

ology to avoid mistakes that cannot be detected by

logical formalisms alone is the formal use of Onto-

logical Realism. We highlight the reasoning power

that formal ontological relations provide to a carefully

crafted ontology given the higher semantic level that

these relations comprise (Smith and Ceusters, 2010).

The formalization of Ontological Relations has been

advocated for many years and it succeeded in the

development of "relations that obtain between enti-

ties in reality, independently of our ways of gaining

knowledge about such entities" (Smith et al., 2005).

OGCP as Suitable Support for Clinical Practice

Knowledge. It is an ontology of entities involved

in a clinical encounter. OGCP includes very general

terms that are used across medical disciplines, includ-

ing: ’disease’, ’disorder’, ’disease course’, ’diagno-

sis’, ’patient’, and ’healthcare provider’. OGCP uses

the Basic Formal Ontology (BFO) (BFO, 2012) as an

upper-level ontology. OGCP provides a formal theory

of disease and treatment. This theory is implemented

using OWL-DL and is available in OWL.

SNOMED CT as the Primary Terminology Aggre-

gation. Our effort will take advantage of the breadth

of coverage of SNOMED CT in our domain of inter-

est(Smith and Brochhausen, 2010).

It has an underlying description logic (EL family).

EL family has shown to be suitable for medical ter-

minology processing and subsequently, E LH R + is

the performance target of many modern classiﬁers in-

cluding those based in consequence driven reasoning

capable of classifying SNOMED CT in practical and

acceptable processing times with recent proposed ex-

tension for concurrent processing (Simancik et al.,

2011) that beneﬁts of current advances in BigData

cluster processing.

OGCP abiding to OBO Foundry. The upper-level

ontologies that support OGCP are introduced in what

regards the ontological relations that can and shall be

used: BFO (IFOMIS, 2004) is strictly focused on the

task of providing a genuine upper ontology which can

be used in support of domain ontologies developed for

scientiﬁc research within the framework of the OBO

Foundry; FMA; RO(RO, 2012) and AIAO.

Clinical Ontology Fine Tuning. Starting from the

ontology alignment in the following ﬁgure.

Figure 1: Ontology Alignment.

We made an effort of trimming and pruning of

the OGMS and CPR complementing in accordance

to our team of cardiologists to better accommodate

their needs expressed in the reports we sampled. That

included some "gardening" to include: SO the Symp-

tom Ontology, VSO the Vital Signs Ontology and oth-

ers all of them accord to OBO Foundry principles. In

order to align the clinical concepts in the various on-

tologies present, an effort was needed to amalgamate

them according to a sound theory of disease and that’s

why we incorporate the DO that was expressly built

with this purpose in mind (Schriml et al., 2012). The

DevelopmentandPopulationofanElaborateFormalOntologyforClinicalPracticeKnowledgeRepresentation

287

Disease Ontology is a community driven, open source

ontology that is designed to link disparate datasets

through disease concepts. It’s provided a computable

structure of inheritable, environmental and infectious

origins of human disease to facilitate the connection

of genetic data, clinical data, and symptoms through

the lens of human disease (Wiki-DO, 2012). The DO

semantically integrates disease and medical vocabu-

laries through extensive cross mapping and integra-

tion of MeSH, ICD, NCI Thesaurus, SNOMED CT

and OMIM (OMIM, 2012) disease-speciﬁc terms and

identiﬁers. It represents a comprehensive knowledge

base of 8043 inherited, developmental and acquired

human diseases.

Ontology Learning From Text. The problem of

acquiring the knowledge necessary for ontology pop-

ulation known as the "Knowledge Acquisition Bottle-

neck" is a challenging (Wong et al., 2012) issue that

remains one of the main barriers for automated acqui-

sition and we tried to circumvent it by using a progres-

sive tutored learning approach. We start from semi-

structured text and use the semi-automated translation

tasks to generate a controlled domain speciﬁc vocab-

ulary on which further acquisition tasks build upon

(Mendes and Rodrigues, 2012b) minimizing ambi-

guity and redundancy for better reasoning capabili-

ties. When instantiating individuals (populate) formal

heavyweight ontologies like the OGCP we do not nor-

mally intend to enrich the ontology but instead turn

them from theoretical models of the domain into rea-

soning able knowledge bases.

4 RESEARCH METHODOLOGY

As reviewed in (Wong et al., 2012) the state-of-the-

Art for acquisition from Clinical Text has enjoyed

strong developments in recent years. We are diving

into extraction from free text present in most inter-

faces used by clinicians. So far we are elaborating in

Cardiovascular related healthcare.

4.1 Text Ontological Annotation

To convert the source texts (clinical note, release re-

port, exam report, or others) the process is a trans-

forming sequence that involves several sequential

steps. This transformation renders ultimately an

OGCP instance. These tasks can be done manually or

automated. Those steps workﬂow can be conﬁgured

declaratively using the software architecture shown in

section 5.1. There are steps involved that consist of:

Figure 2: SOAP Report Sample.

• PDF to raw text or to structured (XML) convert-

ing for adequate documents cleansing. For in-

stance the graphical presentation of Vital Signs

that are originally rendered in the respective report

has to be deleted from the document for easier

terms processing and the tables with values must

be structured accordingly for the annotators to be-

have properly. Initially there is a proof of con-

cept that involves manually cleaning the original

reports

• Manual translation (that is indispensable for the

translator tutoring as shown in 4.2) with the

precise clinicians validation of their jargon ade-

quately translated into English,

• Annotation, either manually using the Web inter-

face of any of the services that we introduce in 5.2,

or automatically through the Web Services avail-

able

• Filtering the concepts from the annotated text to

insert in OGCP instances

Given the array of available Web Services that can se-

mantically annotate bio-medical concepts in English

that are presented ahead in section 5.2, we chose to

use an evolutionary approach for use of the BioPortal

annotator (Noy et al., 2009). We mean by evolution-

ary approach the fact that we ﬁrst use the annotator af-

ter manual pre-processing and then a more automatic

workﬂow.

4.2 Using Automated Translation for

Concept Uniﬁcation

We can take advantage of the fact that we have to

translate from jargon to English to customize the

Google translator toolkit

with our own Translation

https://translate.google.com/toolkit

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

288

Memories and Glossaries. Let us introduce some

demonstrative examples taken from the sample doc-

ument gently provided by Dr. Carlos Baeta and prop-

erly de-identiﬁed presented in the ﬁgure "SOAP Re-

port Sample" above.

We will, in the process of using the Google toolkit,

create Translation Memories with the identiﬁed per-

sonal acronyms like:

• AP (Antecedentes Pessoais) into Personal History

• HTA (Hiper Tensão Arterial) into High Blood

Pressure

• FA (Fibrilhação Auricular) into Atrial Fibrilation

• V. Mitral (Válvula Mitral) into Mitral Valve

Some which are acronyms that can be given the suit-

able translated concept like:

• ECG (Electro Cardio Grama) into Electro Cardio

Gram

or those that are even English acronyms:

• INR (International Normalized Ratio) into Inter-

national Normalized Ratio

Included in this sample are notorious some more com-

plex problems that are not related to the translation it-

self but with some other problems like the time span-

ning of concepts like “1 comp/dia” which is ade-

quately translated to “1 tablet per day” using the de-

ﬁned Translation Memory but has to be posteriorly

exactly characterized as time delimited occurring pro-

cess.

4.3 SOAP Report

This report depicts a clinical encounter in a semi-

structured way. As seen previously in the ﬁgure in

this section we ﬁnd sections that can be associated

with

Symptoms, the subjective section S where we extract

directly to ogcp:symptom-recording.

Signs, the objective section O that are

ogcp:sign-recording that we take as gen-

erator for ogcp:clinical-findings.

Actions, the analysis section A which are the

ogcp:clinical-investigation-act whose outputs

can be ogcp:clinical-artifact to investigate

things that can be ogcp:isConsequenceOf

any of ogcp:physiological-process or

ogcp:pathological-process

and ﬁnally

Plan, the plan section P where the therapeutic acts

can be extracted with all the timing, posology and

prescriptions registered in a particular clinical en-

counter.

We ﬁnd then that if a sufﬁcient amount of clin-

ical reports are fed into the knowledge base it will

ultimately build a sound picture of a clinical practice.

For the inferred axioms to be believable by the com-

munity and thus usable as a Clinical Practice Support-

ing Tool that generated knowledge has to be clearly,

although basically, explained to the user.

4.4 The Complete Acquisition

Workﬂow Picture

Text cleansing

•Lexical, Syntactic and Morphological tagging

•Acronym identification through TM development

Accurate translation

Raw text convert

Expanded Semantic annotation through

BioPortal

Export into TXT, CSV or XML

GRDDL XSLT

OGCP instance creation

Figure 3: Acquisition Flowchart.

The ﬂowchart that depicts graphically the acqui-

sition from the source texts in Portuguese to the cre-

ation of the appropriate OGCP instance is shown in

ﬁgure "Acquisition Flowchart"

5 RESULTS

5.1 Software Implementation

We have presented the full conception of an archi-

tecture in (Mendes and Rodrigues, 2012a). We are

proposing an extensible Enterprise Architecture based

in a ESB which we call CP-ESP (Clinical Practice -

Enterprise Service Bus). This is a common rail where

messaging can ﬂow using a subscription model that

enables the communication to be detached from any

two particular services but instead be available on-

request by one and served by another in a loosely

coupled way. The ESB can then intervene in mes-

sage exchange and overwrite standard rules for ser-

vice execution. The case of an intervention here is

DevelopmentandPopulationofanElaborateFormalOntologyforClinicalPracticeKnowledgeRepresentation

289

the ability to ﬁlter and redirect invocations to the

appropriate NLP task processors depending on the

source being labeled with the status of the load it

carries. The REST philosophy is suggested in our

proposal as the best way of implementing a Service

Oriented Architecture that serves as the communica-

tion underlying structure of our system. REST end-

points are available for the generality of our needs.

The available Web Services can render responses in

highly-structured forms like JSON or in any of the

standardized mime types that can be handled by the

ﬁltering and enqueuing capabilities of any conﬁg-

urable available ESB like those based in Apache Ser-

viceMix (http://servicemix.apache.org/home.html) or

Mule (http://www.mulesoft.org/) for instance. They

can be conﬁgured to compose a complete pipeline

very easily:

Figure 4: Instance Translation and Creation.

In the picture is shown the possibility of deﬁning

services (daemons) that monitor the presence of new

reports and appropriately apply the needed transfor-

mations according to their status and content until a

translated document is delivered to the adequate end-

point for annotation. All of this can even be done with

due care about scalability, availability and all prob-

lems associated with a modern state-of-the-art soft-

ware architecture as presented in (Mendes and Ro-

drigues, 2012a).

In the picture is shown the possibility of deﬁn-

ing services (daemons) that monitor the presence

of new reports and appropriately apply the needed

transformations according to their status and con-

tent until a translated document is delivered to the

adequate end-point for annotation. All of this can

even be done with due care about scalability, avail-

ability and all problems associated with a modern

state-of-the-art software architecture as presented in

(Mendes and Rodrigues, 2012a) Building over the

suggested infrastructure the systems are rather com-

posed as opposed to monolithically built and so man-

ifest high capabilities of plug-and-play conﬁguration

allowing for interchangeable providers (as Web Ser-

vices), Reference Ontologies (Feeders), and target

ontologies. Having the foundations available with

the right weapons provided one has to take a prac-

tical approach to the development of a target system

using, in our case, the Java best-practices for prag-

matic development that include a number of Patterns

as in JEE (Java Enterprise Edition) or the pragmatic

approaches developed in such successful projects as

Spring (http://www.springsource.com/).

5.2 OGCP-ESB

We use, through BioPortal and ODIE, algorithms for

Named Entity Recognition, Co-reference resolution,

concept discovery, discourse reasoning and attribute

value extraction. A local version gives the possibility

of developing extensions to the algorithms provided

in the base offering allowing, for instance, targeting

different languages in the NLP tasks. The Web Ser-

vices provided by BioPortal or ODIE can be locally

extended and reﬁned for different sources and are pro-

vided as one of the project deliverables.

5.3 NLP Pragmatics and Discourse

Controller

The text for any particular encounter (actually for any

Clinical Episode) may be collected in the form suit-

able for processing into the Ontology framework us-

ing some NLP pragmatics. Populating the OGCP

the “Clinical Picture” is completed and thus our KB

is available for validation and further logical infer-

encing. The semantic representation is done us-

ing pragmatic interpretation as deﬁned in our fellow

researcher at CENTRIA Dora Melo’s article (Melo

et al., 2012). The enrichment process must always

maintain the entailments provided by the base (gold-

standard) ontologies and so can never lead to incon-

sistency. We use a round-trip, debug and repair, build-

ing method to populate/enhance the OGCP then. For

any new instance the validation is performed and new

possible inferred facts generated if consistency is yet

valid. These new facts are candidates for NLP justiﬁ-

cations generation. The main objective of the system

is to provide accurate answers to questions posed by

users and, in our proposal these answers are clinically

valid because the generation method guarantees that.

QA is, however, only one of the interesting features

of our work that is enhanced by the adequate justi-

ﬁcation to be evidently useful for practitioners. To

develop justiﬁcations from DL

arguments inferred

from consequence based reasoners (Kazakov, 2009;

Description Logic

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

290

Kazakov et al., 2011) we based our work in (Bail

et al., 2011) to study and compare the justiﬁcatory

structure to those present in the NCBO BioPortal ad-

dressed in the mentioned article. The results so far

are in the realm of ’ontology verbalization’, the gen-

erated explanations are still in a controlled natural

language (CNL) fashion. The obtained results seem

to be adequate enough for the users to ﬁnd them be-

lievable and thus the justiﬁcations stand in our con-

trolled clinical setting. We use the verbalization tool-

ing (Kaljurand, 2010) to present the justiﬁcations in

an acceptable manner. The foundational techniques

were introduced in (Kaljurand and Fuchs, 2007). For

the verbalization to function properly all the restric-

tions of content are guaranteed in the process of on-

tology (Knowledge base) enrichment from SOAP re-

ports. For instance, all names are English words and

individuals are singular proper names (preferably cap-

italized) named classes are denoted by singular count-

able nouns and (object) properties by transitive verbs

in their lemma form (i.e. inﬁnitive form) (Kalju-

rand and Fuchs, 2007). The decision of what inferred

knowledge is then presented with its justiﬁcations to

the user is a task handled by the DC

using the de-

veloped pragmatics introduced in the above referred

article (Melo et al., 2012). In our proposed methodol-

ogy we make use os the DC to align the most relevant

clinical terms into a acceptable CNL document and

the process may be graphically presented as:

Figure 5: Discourse Controller Flow.

6 CONCLUSIONS

We propose an OGMS extension using the adequate

ontological realism approaches and incorporating the

CPR and its upper level ontologies as framework for

an EL reasoning workhorse. We present our efforts

for knowledge base population from semi-structured

clinical text reports and discuss the underlying prob-

Discourse Controller

lem of automatic instance creation from them into the

proposed knowledge representation structure. Cur-

rently, after having passed an internal process of

validating the proposed framework (Ontology struc-

ture+Population mechanism) we are assessing its ac-

ceptance in a wider regional level by exposing a group

of related cardiologists to the works done so far and

receiving their feedback in a formal and traceable

manner. We are developing a knowledge represen-

tation infrastructure enabling the usage of highly op-

timized distributed consequence based reasoners that

are referred in literature only in 2011. With these very

recent developments it’s ﬁnally possible to validate

the enormous knowledge bases that are created by au-

tomatically populating the proposed ontology OGCP

that relies on extensive, and very solid, foundations

like SNOMED-CT and FMA among others. Logi-

cal inferencing and clinical facts entailment that is

possible through this capability is an interesting con-

tribution to the application of Artiﬁcial Intelligence

to healthcare. We introduce clinical decision sup-

port systems (CDSS) that are based on such a break-

through technique. We further argue that it is impera-

tive, for the broad acceptance of these tooling by the

medical community, that their inferences are justiﬁed

using controlled natural language and adequate termi-

nology.

7 FUTURE WORK

We intend to deliver OGCP in the NCBO BioPor-

tal as soon as it reaches a minimum of usability as

KR tool for cardiology which we intend to happen

later this year. We are currently in the evaluation pro-

cess mentioned above and, depending on the level of

usability/acceptance, will evolve to different clinical

specialties to demonstrate its ﬂexibility. Shall all the

efforts prove to be worthy we will try to extend it

to wider levels following a community development

process based in some OSS (Open Source Software)

repository like Google Code for example. It will all

be summed up in the PhD thesis of the ﬁrst author to

be presented late 2013.

ACKNOWLEDGEMENTS

We acknowledge IIFA - Instituto de Investigação e

Formação Avançada of Universidade de Évora for the

Bento de Jesus Caraça scholarship that is awarded to

the ﬁrst author and CENTRIA Center for Research in

Artiﬁcial Intelligence of Faculdade de Ciências e Tec-

DevelopmentandPopulationofanElaborateFormalOntologyforClinicalPracticeKnowledgeRepresentation

291

nologia of Universidade Nova de Lisboa for its

continuous ﬁnancial support of our work.

REFERENCES

Bail, S., Horridge, M., Parsia, B., and Sattler, U. (2011).

The justiﬁcatory structure of the ncbo bioportal on-

tologies. In Aroyo, L., Welty, C., Alani, H., Taylor, J.,

Bernstein, A., Kagal, L., Noy, N. F., and Blomqvist,

E., editors, International Semantic Web Conference

(1), volume 7031 of Lecture Notes in Computer Sci-

ence, page 67Ã

câ

nâ

AIJ82. Springer.

BFO (2012). Basic formal ontology 2.0.

Brochhausen, M., Burgun, A., Ceusters, W., Hasman, A.,

Leong, T., Musen, M., Oliveira, J., Peleg, M., Rector,

A., Schulz, S., et al. (2011). Discussion of" biomedi-

cal ontologies: Toward scientiﬁc debate". Methods of

information in medicine, 50(3):217.

Hoehndorf, R., Dumontier, M., Oellrich, A., Rebholz-

Schuhmann, D., Schoﬁeld, P., and Gkoutos, G.

(2011). Interoperability between biomedical ontolo-

gies through relation expansion, upper-level ontolo-

gies and automatic reasoning. PloS one, 6(7):e22006.

IFOMIS (2004). Basic formal ontology.

Kaljurand, K. (2010). Owl verbalizer: making machine-

readable knowledge also human-readable. Available

Online: https://code.google.com/p/owlverbalizer/.

Kaljurand, K. and Fuchs, N. E. (2007). Verbalizing

owl in attempto controlled english. Proceedings of

OWLED07.

Kazakov, Y. (2009). Consequence-Driven Reasoning for

Horn SHIQ Ontologies. In Boutilier, C., editor, IJCAI,

pages 2040–2045. IJCAI Distinguished Paper Award

Winner.

Kazakov, Y., Krötzsch, M., and Simancík, F. (2011). Con-

current classiﬁcation of el ontologies. In Proceedings

of the 10th international conference on The seman-

tic web - Volume Part I, ISWC’11, pages 305–320,

Berlin, Heidelberg. Springer-Verlag.

Melo, D., Rodrigues, I., and Nogueira, V. (2012). Work

out the semantic web search: The cooperative way.

Advances in Artiﬁcial Intelligence, 2012.

Mendes, D. and Rodrigues, I. P. (2012a). A Semantic

Web pragmatic approach to develop Clinical Ontolo-

gies, and thus Semantic Interoperability, based in HL7

v2.xml messaging. In Information Systems and Tech-

nologies for Enhancing Health and Social Care. IGI

Global.

Mendes, D. and Rodrigues, I. P. (2012b). Advances to se-

mantic interoperability through cpr ontology extract-

ing from soap framework reports. electronic Journal

of Health Informatics.

Noy, N. F., Shah, N. H., Whetzel, P. L., Dai, B., Dorf, M.,

Grifﬁth, N., Jonquet, C., Rubin, D. L., Storey, M.-A.,

Chute, C. G., and Musen, M. a. (2009). BioPortal:

ontologies and integrated data resources at the click

of a mouse. Nucleic acids research, 37(Web Server

issue):W170–3.

Ogbuji, C. (2011). A Framework Ontology for Computer-

Based Patient Record Systems. In Proceedings of the

ICBO: International Conference on Biomedical On-

tology, pages 217–223, Buffalo, NY, USA.

OGMS (2010). Ontology for general medical science.

OMIM (2012). Omim, online mendelian inheritance in

man. Available Online: http://www.ncbi.nlm.nih.gov/

omim.

RO, O. F. (2012). Relations ontology.

Schriml, L., Arze, C., Nadendla, S., Chang, Y., Mazaitis,

M., Felix, V., Feng, G., and Kibbe, W. (2012). Dis-

ease ontology: a backbone for disease semantic inte-

gration. Nucleic Acids Research, 40(D1):D940–D946.

Simancik, F., Kazakov, Y., and Horrocks, I. (2011).

Consequence-Based Reasoning beyond Horn Ontolo-

gies. In Walsh, T., editor, IJCAI, pages 1093–1098.

IJCAI/AAAI.

Smith, B. and Brochhausen, M. (2010). Putting Biomedical

Ontologies to Work. Methods Inf Med, 49.

Smith, B. and Ceusters, W. (2010). Ontological realism:

A methodology for coordinated evolution of scientiﬁc

ontologies. Applied ontology, 5(3-4):139–188.

Smith, B., Ceusters, W., Klagges, B., KÃ˝uhler, J., Kumar,

A., Lomax, J., Mungall, C., Neuhaus, F., Rector, A. L.,

and Rosse, C. (2005). Relations in biomedical ontolo-

gies. Genome Biol, 6(5):R46.

W3C (2009). Computer-based patient record ontology.

http://code.google.com/p/cprontology. http://code.

google.com/p/cprontology.

Wiki-DO (2012). Disease ontology wiki page. Avail-

able Online: http://do-wiki.nubic.northwestern.edu/

do-wiki/index.php/Main_Page.

Wong, W., Liu, W., and Bennamoun, M. (2012). Ontology

learning from text: A look back and into the future.

ACM Comput. Surv., 44(4):20:1–20:36.

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

292