BFO-Based Ontological Analysis Framework: Definition Annotation Manual (Version 1)

1/16 BFO-Based Ontological Analysis Framework Definition Annotation Manual (Version 1) By Selja Seppälä Last modified: October 8, 2015 Introduction This definition annotation manual provides guidelines for: • Segmenting natural language definitions into their genus and differentia parts. • Tagging each segment with an ontological label provided by a relational model that is based on any of the categories of the Basic Formal Ontology (BFO), a realist upper-level ontology (Arp, 2015; Smith et al., 2015; Spear, 2006). This manual specifies the basic principles for structuring a corpus of terminological definitions and for implementing the proposed BFO-based ontological annotation schema in XML. Note These specifications correspond to the first version of the annotation schema. The schema is currently being updated to allow, among other things, for an easier integration with OWL. Background This definition annotation schema is part of the 'BFO-Based Ontological Analysis Framework' put forward in Seppälä (2012, 2015a) for carrying out content analyses in terminology. It is an adaptation of the annotation schema defined in Seppälä (2004, chapter IV); both are inspired by the conceptual analysis schema put forward by Sager and L'Homme (1994). The ontological analysis framework is in many respects similar to the formalisms adopted by García de Quesada et al. (2002), Montero Martínez (2002), or Montero Martínez et al. (2002) in their treatment of the terminology of oncology. The guiding principle in all these works is that definitions are segmented in genus and differentia elements, and then tagged with conceptual 'class' and 'relation' labels. The main DefinitionAnnotationManual Introduction 2/16 difference with these terminological annotation frameworks is that the set of descriptors proposed here does not represent concepts but types of entities in the world, their properties, and relations between them. This difference brings about distinct choices regarding the annotation schema, especially regarding the processing of the genus element. Another major difference is that the categories and relations used in the ontological analysis framework are very general, as the Basic Formal Ontology represents domain-independent entity types and relations between them. Definitions What is meant here by definition is the short text that can be found in the 'definition' field of a (specialized) dictionary entry or in an annotation field of an ontology. Roughly speaking, it generally corresponds to a single sentence that gives a description of the meaning of a term and that does not include the term itself. The following examples illustrate terminological definitions from the domains of chemical weapons (Aureille, 2003) and oil spill cleanup (Graf, 2003). distilled mustard sulfur mustard bis (2-chloroethyl) sulfide A blister agent which is an amber brown liquid with an odor similar to that of burning garlic. submunition bomblet A chemical munition of small size contained in a main one and designed to disperse non-persistent agents. net boom A boom that is made of netting to facilitate the retention of viscous oils. pump system The part of the skimmer that transfers the recovered oil, oil and water and/or emulsions from the skimmer head to a storage tank. Relational Models What is meant here by relational model is a template that specifies the 'ENTITY TYPE+relation+ENTITY TYPE' triples, called relational configurations (RCs), that characterize each category of BFO1. Each model consists of the RCs that characterize DefinitionAnnotationManual Introduction 3/16 each category, as well as the RCs that these categories inherit from their parent categories in the ontology – since BFO categories are hierarchically organized through 'is_a' relations. The following example shows the relational model for the category OBJECT in BFO 2.0, the first version of the ontology2. OBJECT Relational configurations characterizing the entity type OBJECT <has_part object> <has_part object aggregate> Relational configurations inherited from higher levels <bearer_of disposition> <bearer_of quality> <contains process> <contains process boundary> <has_history process> <has_part immaterial entity> <has_part material entity> <located_in independent continuant> <material_basis_of disposition> <occupies three-dimensional spatial region> <part_of immaterial entity> <part_of material entity> <participates_in process> These relational models are intended to describe and predict the contents of the definitions according to the category of the thing to which the defined term – and thus the definition itself – refers. Internal Structure of Definitions This ontological annotation schema is intended to reveal the internal structure of (terminological) definitions in terms of ontological categories, and relations between them. The following examples show a possible output of the above definitions annotated using this annotation schema and the model for the category OBJECT. distilled mustard sulfur mustard DefinitionAnnotationManual Introduction 4/16 bis (2-chloroethyl) sulfide REF <is_a object> GEN <bearer_of role> A blister agent SPE <bearer_of quality> which is an amber brown liquid with an odor similar to that of burning garlic. submunition bomblet GEN <is_a object> A chemical munition SPE <bearer_of quality> of small size SPE <located_in site> contained in a main one SPE <bearer_of function> and designed to disperse non-persistent agents. net boom GEN <is_a object> A boom SPE <has_part object> that is made of netting SPE <bearer_of function> to facilitate the retention of viscous oils. pump system REF <is_a object> GEN <part_of object> The part of the skimmer SPE <bearer_of function> that transfers the recovered oil, oil and water and/or emulsions from the skimmer head to a storage tank. Usefulness of the BFO-based Ontological Annotation Framework Corpora of existing definitions annotated with this schema can, for example, be analyzed in order to test the hypothesis according to which the contents of definitions are at least in part influenced by the type of entity to which the referent of the defined term belongs (Seppälä, 2012, 2015a). If this hypothesis is verified, then it is possible to use the annotated definitions to see to what extent the entity types influence the defining contents. The annotated corpus can also yield statistical information on the kind of RCs that are relevant for defining terms, thus revealing typical RCs for each BFO category. The statistical results can also be used to assign weights to the RCs in the models. 5/16 Specifications of the Annotation Schema The annotation of definitions comprises two sub-tasks that can be realized separately or simultaneously: 1. Segmenting the definition into two or more parts. 2. Tagging the parts with an ontological label provided by the 'relational models' that are based on the upper-level realist Basic Formal Ontology (BFO). The annotations of the definitions are done in XML. The annotation schema contains general tags that identify the dictionary entries constituting the corpus to be analyzed, the domain and sub-domain(s) to which they pertain, and their different fields: definition, terms, notes, phraseological information, and references to other entries. The ontological analysis as such only applies to the content of the definition field, that is, the defining sentence in between the <DF> tags. The following example shows an excerpt of a corpus of multi-domain definitions structured with these XML tags3. <DEFINITIONS_TERMINOLOGIQUES> <FICHE langue="en"> <NI>EN_MA_111</NI> <CM> <DOMAINE>nettoyage des déversements d'hydrocarbures</DOMAINE> <SS-DOM1>lutte en mer</SS-DOM1> <SS-DOM2>récupération</SS-DOM2> </CM> <VE>pump system</VE> <DF> <GEN typeREF="is_a OBJECT" REFrelGEN="part_of OBJECT">The part of the skimmer </GEN> <SPE REFrelSPE="bearer_of FUNCTION">that transfers the recovered oil, oil and water and/or emulsions from the skimmer head to a storage tank. </SPE> </DF> <PH>Based on past experience, the chief cause for an unsuccessful oil skimming operation has been due to failure in the pump system.</PH> </FICHE> <FICHE langue="en"> <NI>EN_SS_151</NI> <CM> <DOMAINE>natation synchronisée</DOMAINE> <SS-DOM1>mouvement</SS-DOM1> <SS-DOM2>sortie</SS-DOM2> </CM> <VE>walkout</VE> <DF> <GEN REFrelGEN="is_a PROCESS">Movement </GEN> <SPE REFrelSPE="has_part PROCESS">which starts in a split position, unless DefinitionAnnotationManual Annotation Schema 6/16 otherwise specified in the figure description, in which the hips remain stationary as one leg is lifted in an arc over the surface to meet the opposite leg.</SPE> </DF> <NT/> <PH/> </FICHE> .. </DEFINITIONS_TERMINOLOGIQUES> These structuring tags are explained in the following sections. Structuring)the)Corpus This%sec(on%presents%the%XML%elements%and,%if%applicable,%their%a8ributes%for%structuring%the% corpus%of%defini(ons%to%be%analyzed. <DEFINITIONS_TERMINOLOGIQUES>:% 'terminological_defini(ons'% element% that% sets% the% beginning%and%end%of%the%corpus%of%defini(ons. <FICHE)langue="fr">:%'record'%element%that%sets%the%beginning%and%end%of%a%terminological% entry%or%record.%This%element%has%an%a8ribute%'langue'%for%specifying%the%language%of%the%record% with%the%ISO%639H1%twoHle8er%language%codes%(e.g.,%’en'%for%English;%'fr'%for%French). <NI>:% 'id'% element% that% sets% the% beginning% and% end% of% a% unique% iden(fier% for% each% FICHE% element4.% <CM>:% 'subjectHma8er'% element% that% sets% the% beginning% and% end% of% the% domain% and% subH domain%specifica(ons5. <DOMAINE>:% 'domain'% element% that% sets% the% beginning% and% end% of% the% main% domain% to% which%the%terminological%record%pertains. <SSHDOM#>:%'subHdomain'%element%that%sets%the%beginning%and%end%of%the%subHdomain(s)%to% which%the%terminological%record% pertains6.%Since% there%can%be%several%hierarchically%organized% subHdomains,% this% element% includes% a% number% ('#')% specifying% the% level% in% the% hierarchy% (<SSH DOM1>%to%<SSHDOM3>). <VE>:%'term'%element%that%sets%the%beginning%and%end%of%a%term7. DefinitionAnnotationManual Annotation Schema 7/16 <DF>:% 'defini(on'% element% that% sets% the% beginning% and% end% of% the% defini(on.% The% <DF>% element% includes% subHelements% that% set% the% beginning% and% end% of% the% different% parts% of% the% defini(on%(<GEN>%and%<SPE>%elements),%thus%marking%the%internal%structure%of%the%defini(ons. <GEN>:%'genus'%element%that%sets%the%beginning%and%end%of%the%genus%element8. <SPE>:%'specifier'%(or%'differen(a')%element%that%sets%the%beginning%and%end%of%a%specif%ier%(or% differen(a)%element9. <NT>:%'note'%element%that%sets%the%beginning%and%end%of%a%linguis(c%or%encyclopedic%note.% <PH>:% 'phraseology'% element% that% sets% the% beginning% and% end% of% some% phraseological% (linguis(c)%informa(on%or%example. <REL>:%'related%term'%element%that%sets%the%beginning%and%end%of%a%crossH1reference%to%one% or%more%other%terms%in%the%domain. The% <GEN>% and% <SPE>% elements% include% a8ributes% for% the% ontological% tagging% of% each% segment,% respec(vely,% with% the% categories% of% the% rela(onal% models,% and% their% RCs.% These% elements%are%presented%in%more%detail%in%the%following%sec(ons. SegmenMng)the)DefiniMons Defini(ons%are%segmented%in%parts%that%fulfill%two%different%roles%within%the%defini(on:%the% genus,%and%one%or%more%specifier%(or%differen-a)%elements. The)Genus)Element A%defini(on%has%only%one%genus%element%(GEN)%that%serves%to%categorize%the%thing%to%which% the%defini(on%refers%–%which%is%also%the%referent%of%the%defined%term.% The%genus%of%a%defini(on%is%enclosed%within%the%following%opening%and%closing%tags: <GEN>Genus element</GEN> The%genus%part%of%the%defini(on%normally%relates%the%defined%term%to%a%more%general%one.% The%genus%can%be%a%term%of%the%same%domain%as%the%defined%one%or%a%common%word.%When%it%is% a%term%of%the%domain,%it%can%be%an%immediate%parent%of%the%defined%term%(the%genus/proximus)% or%a%more% general% parent% term.%When% the% term%is%the%genus/proximus,%it%is%marked% with% the% following%a8ribute10: DefinitionAnnotationManual Annotation Schema 8/16 <GEN relationVE="GENRE_PROCH">Genus element</GEN> The)Genus)and)the)Defined)Term)Have)the)Same)Referent Normally,%the%genus%part%of%the%defini(on%expresses%the%categorizing%'is_a'%rela(on%(although% the%rela(on%is%only%implicit).%In%this%case,%the%genus%refers%to%the%same%category%as%the%referent% of% the% defined% term.% This% is% reflected% in% the% 'is_a% CATEGORY'% value% of% the% a8ribute% REFrelGEN11: <GEN REFrelGEN="is_a CATEGORY" relationVE="GENRE_PROCH">Genus element</GEN> The)Defined)Term)is)an)Instance)of)the)Referent's)Category Some(mes,%the%genus%part%of%the%defini(on%expresses%an%'instance_of'%rela(on,%meaning% that%the%defined%term%is%a%par(cular%instance%of%the%more%general%category%expressed%in%the% GEN.%For%example,%the/Large/Hadron/Collider%or%LHC%is%an%instance%of%a%par-cle/accelerator.%In% this%case,%the%value%of%the%a8ribute%REFrelGEN%is%as%follows: <GEN REFrelGEN="instance_of CATEGORY">Genus element</GEN> ParMcular)cases When%segmen(ng%and%tagging%defini(ons,%the%annotator%might%however%find%different%cases% that%depart%from%the%ideal%case%where%the%GEN%tag%states%the%category%of%the%defined%term's% referent.% The)Referent)of)the)Genus)and)of)the)Defined)Term)Differ Some(mes,%the%genus%part%of%the%defini(on%refers%to%a%category%that%is%different%from%the% category%to%which%the%defined%term%refers.%In%these%cases,%the%referent%of%the%defined%term%bears% some%other%rela(on%to%the%category%expressed%in%the%genus,%such%as%'part_of'.%The%ontological% analysis% framework% addresses% these% cases% explicitly% by% specifying% both% rela(ons% in% separate% a8ributes: • The% 'is_a'% or% 'instance_of'% rela(on% between% the% referent% of% the% defined% term% and% the% category%to%which%it%belongs%(see%the%typeREF%a8ribute).% • Another%rela(on%between%the%referent%of%the%defined%term%and%the%category%expressed% by%the%genus%(see%the%REFrelGEN%a8ribute). <GEN typeREF="is_a CATEGORY" REFrelGEN="other_relation CATEGORY">Genus element</GEN> The)DefiniMon)Has)No)Part)Fulfilling)the)Genus)Role Some(mes,% the% defini(on% has% no% genus% part% at% all% (for% example% in% defini(ons% of% adjec(val% terms).%In%these%cases,%an%empty%GEN%tag%with%the%typeREF%a8ribute%is%added%before%the%(first)% SPE%tag: DefinitionAnnotationManual Annotation Schema 9/16 <GEN typeREF="is_a CATEGORY"/><SPE>Some specific element</SPE> ATributes)of)the)<GEN>)element) As%we%have%seen,%the%ontological%analysis%framework%provides%a%set%of%XML%a8ributes%to%the% <GEN>% element% that% address% each% of% these% cases% while% yielding% an% explicit,% consistent,% and% uniform%annota(on%of%the%defini(ons.% The%annota(on%schema%also%provides%a8ributes%for%cases%where%the%genus%element%is%cut%by% a% SPE% element% (voir_GEN)% and% for% indica(ng% categoriza(on% issues% (statutTypeREF% and%% statutREFrelGEN).% The%<GEN>%element%with%all%its%a8ributes%looks%like%this: <GEN typeREF="is_a|instance_of CATEGORY" REFrelGEN="some_relation CATEGORY" relationVE="GENRE_PROCH" voir_GEN="DEBUT|FIN" statutTypeREF="categ" statutREFrelGEN="categ"> typeREF:%a8ribute%that%indicates%the%category%to%which%the%defined%term%refers%when%it%is% different%from%the%category%to%which%the%GEN%refers%or%when%the%GEN%tag%is%empty.%Its%value%is% always% composed% of% an% 'is_a'% or% 'instance_of'% rela(on% followed% by% a% category.% REFrelGEN:% a8ribute%that%indicates%the%rela(on%between%the%referent%of%the%defined%term%and%the%category% expressed%by%the%genus%element%(in%that%order).%Its%value%is%always%composed%of%some%rela(on% followed%by%a%category. relaMonVE:%a8ribute%that%indicates%that%the%category%in%the%GEN%is%the%genus/proximus,%i.e.,% the%immediate%parent%category%of%the%defined%term.%This%a8ribute%is%included%only%when% applicable%and%its%value%is%GENRE_PROCH. voir_GEN:%a8ribute%that%indicates%that%the%GEN%is%split%in%two%parts%by%a%SPE%element.%It%has% the%values%DEBUT%for%the%beginning%of%the%GEN%and%FIN%for%the%last%part%of%the%GEN. statutTypeREF:%a8ribute%that%indicates%that%the%categoriza(on%of%the%RC%specified%in%the% typeREF%tag%is%uncertain.%This%a8ribute%is%included%only%when%applicable%and%its%value%is%categ. statutREFrelGEN:%a8ribute%that%indicates%that%the%categoriza(on%of%the%RC%specified%in%the% REFrelGEN%tag%is%uncertain.%This%a8ribute%is%included%only%when%applicable%and%its%value%is%categ. The Specifier Elements (Differentiae) A specifier (or differentia) element (SPE) serves to specify the information provided by the genus thus narrowing down the meaning of the defined term. In the ideal case, a specifier differentiates the defined term from the genus and from its neighboring terms – DefinitionAnnotationManual References 16/16 7 'VE'%stands%for%'vede8e'%in%French,%that%is%'headword'%or%'entry%term'%in%English. 8 The%genus%element%is%further%defined%in%the%corresponding%sec(on%(see%page%7). 9 The%specifier%(or%differen(a)%element%is%further%defined%in%the%corresponding%sec(on%(see% page%9). 10 'VE'%stands%for%'vede8e',%the%French%term%for%'entry'%in%a%dic(onary.%Thus,%'rela(onVE'%is%to% be%understood%as%the%rela(on%between%the%genus%term%of%the%defini(on%and%the%defined%term,% as%in:%'The%genus%term%is%the%genus/proximus%of%the%defined%term.' 11 As%it%is%reflected%in%the%name%of%the%a8ribute%tag,%all%the%rela(ons%are%unidirec(onal%and% they%are%to%be%understood%as%going%from%the%referent%of%the%defined%term%to%the%category%to% which%the%segment%refers.%This%applies%both%for%GENs%and%SPEs.