Faber, Pamela, Pilar León Araúz, Juan Antonio Prieto Velasco, and Arianne Reimerink. 2007. Linking Images and Words: the description of specialized concepts. International Journal of Lexicography 20, no. 1: 39-65. doi:10.1093/ijl/ecl038.
This is the final peer-reviewed version of the following article: Faber, Pamela, Pilar León Araúz, Juan Antonio Prieto Velasco, and Arianne Reimerink. 2007. Linking Images andWords: the Description of Specialized Concepts. International Journal of Lexicography 20, no. 1:39-65. <http://dx.doi.org/10.1093/ijl/ecl038> You can find more articles authored by LexiCon Research Group members at <http://lexicon.ugr.es>. Linking Images and Words: the Description of Specialized Concepts1 PAMELA FABER PILARLEÓNARAÚZJUANANTONIOPRIETOVELASCO ARIANNEREIMERINK2 University of Granada Abstract A crucial issuein terminology management is how specialized concepts should be represented so asto provide the user with an adequate understanding of their meaning as well as sufficient knowledge of their location within the general knowledge structure of a scientific or technical domain. Sucha conceptualrepresentation should contain information in various formats. In this regard, linguistic and graphical descriptionsof specialized entities play a major role in knowledge representation, especially when both converge to highlight the multidimensional nature of concepts as well as the conceptual relations within a specialized domain. In this article, we explore the nature of thelinks between the linguistic and graphical description of specialized concepts. In a multimodal conceptual description, we believe that thestructured information interminographic definitions should mesh with the visual information in images for a better understanding of complex and dynamic concept systems. 1. Introduction Traditionally, the difference between lexicographyand terminography, apartfrom the fact that the former studies general language and the latter specializedfields of knowledge, hasbeen their methodological approach. Generally speaking, most lexicographic products arestructured alphabetically, whereas terminographic resources tend to have a more conceptual organization. Nevertheless, as Bowker (2003: 155) points out, both lexicographers and terminologists work in a way that often combines elements of semasiological (form-based) and onomasiological(concept-based) approaches.In thelast ten years or so specialized language and general language studies have begun toshowan approximationin perspective aslexicographers andlexical semanticists now tendto takea more concept-based approach towards dictionarystructure, and terminographers havebegun to pay more attention to conceptual description or to the definition of terminological units (Faber et al. 2001, Faber etal. 2005, Faber et al. 2006). Linking Images and Words: the Description of Specialized Concepts One example of this approximation is Frame Semantics (Fillmore 1976, 1982, 1985,Fillmore et al. 2003, Atkins et al. 2003). According to Fillmoreand Atkins (1998), a ‘frame’ canbe defined asanetworkof concepts relatedinsuch a way that one concept evokes the entire system. Frame Semantics,like terminography,isbasedonconceptual organization. In lexicography, frame-based dictionaries have been recently proposed by Boas (2005)and Martin (2006). Frame Semantics, as a theoretical framework, is also consistent with recent research incognitive neuroscience. Barsalou (2003) argues that the human conceptual system isorganized according to situated simulations which act as an interfacebetweencognitionand perception. This situated, dynamic conceptual system is neither fully modal nor fullyamodal and at the highest level oforganization, categories arecreated to streamline the action-environment interface. The notion offrame provides an excellenttool torepresentthe dynamicprocesses that takeplace in specialized fields of knowledgein a way thatissimilar to human conceptual representation. Terminography focuses on the representation of specialized concepts, and morespecifically, on the macro and micro-structural design oftermentries. Since the ultimatepurpose of any terminological resource is to facilitate and enhance knowledge acquisition,the information contained in term entries should be internally as well as externally coherent. Coherence signifies the harmonious flow ofinformation, cooperation, and order among the components of a larger entity, which in this case would be the knowledge resource. In this regard, internal coherence refersto theinformation contained inthedata fields of each entry, whereas external coherence refers to how entries are interrelatedwithin the context of aunified whole. The wider set of options available for linguistic resources in electronic format has generated a corresponding debate regarding how specialized concepts should be represented so as to provide the user with an adequate understanding of a particularconcept as well as sufficient knowledge of its location within the general conceptual network ofascientific ortechnical domain. Depending on the natureof the concept,itisoften useful for such a representation to include information in both linguistic and graphical formats. It is our assertion that linguistic and graphical descriptions ofspecialized entities play a major role in the understanding process when both types of description converge to highlight the multidimensional natureofconcepts3as well as the set of conceptual relations typical of aspecific subdomain. Inthis sense, multidimensionalrefers to the various ways of describing concepts as well as aspecialized domain, using both hierarchical and non-hierarchical relations. 2. Describing concepts We are currently engaged in a research project on knowledge representation in the domain of Coastal Engineering4. One of the products contemplated is a multimedia terminological knowledge base for this specialized field. The elaboration of thisterminological resource has caused us to explore in some depth the nature of the linksbetween the linguistic and graphical description of concepts. However, in a multimodal Linking Images and Words: the Description of Specialized Concepts conceptual description, notjustanyresource willdo since thestructuredinformation in terminographic definitions should mesh with the visualinformation in images in order to provide a deeper understanding of complex and dynamic concept systems. 2.1 The linguistic description of concepts: terminographic definition According to Béjoint (1997: 19–20) definitions have never been given due importance in terminology. In many terminological databases, definitions are simply inserted in a cutand-paste fashion from other dictionaries, term bases, or knowledge resources, without takinginto considerationboththeir internalandexternal coherence. However, definitionsare mini-knowledge representations or frames, and accordingly, the organization ofinformation encoded in definitions should be structured in regard to its perceptual salience as well as its relationship toinformationconfigurationsin the definitions ofother related concepts within the same category (Faber et al. 2001, Faber 2002). Martin (1998)underlines the fact that frames as definition models offer more consistent, flexible, andcomplete representations.When theinformationinsuch representations is enhanced withimages that highlight the different types ofconceptualinformation within each definition, the entry is substantially enriched. According to Grinev and Klepalchenko (1999), the description of specialized domains can in many cases be based on the events that generally take place in them, and can berepresented accordingly. Within our coastal engineering knowledge base, concepts areorganized according to their conceptual category. Although there are various ways of presenting the conceptualizationofa specialized domain, depending ontheobjectives ofthe knowledge resource, we have configured the most generic categoriesin a prototypicaldomain event or action-environment interface (Barsalou 2003: 513, Faber et al. 2005), which providesaframe for the organization of morespecific concepts. The Coastal Engineering Event DESCRIPTION TEMPLATE Attributes of Measurement Disciplines Instruments for Procedures of Representation of Simulation of Prediction of speed, height of time, space for study of description of description of Figure 1. Coastal Engineering Event. Linking Images and Words: the Description of Specialized Concepts As showninFigure 1,ourCoastalEngineeringEvent(CEE)is conceptualizedas a dynamic process that is initiated by an agent (either natural or human), affects a specific kind of patient(a coastal entity),andproduces a result. These macro-categories (AGENT . PROCESS . PATIENT/RESULT)are the concept roles characteristic ofthis specialized domain, and the CEE provides a model to represent their interrelationships. Additionally, there are peripheral categories which include INSTRUMENTSthataretypically usedduring the CEE, as well as acategory where the concepts of measurement, analysis, and description of the processes in the main event are included. The morespecific concepts within eachcategory are organized in a hierarchical network where they are linked by both vertical (hierarchical) and horizontal (non-hierarchical)relations.Inthe database, these concepts canbeaccessedby clicking the hyperlink in the relevant macrocategory. In the following sections, we use concepts representing very diverse entities(i.e.groyne,beach nourishment anddredge)to show how linguistic and visual descriptions in each termentryreflectconceptual structure and relations. 2.1.1 Groyne and beach nourishment. In coastal engineering, a groyne is a wooden, concrete and/or rock barrier at right angles to the sea, which isoftenused to help prevent erosion along the coastline. As such, it is a maritime structure created by a natural or human agent. One of the activities that a human agent can carry out is CONSTRUCTION (artificial process). Construction results in amaritime structure, which is the category that groyne and other similar hard coastal defence structures belong to. On the other hand, beach nourishment designatestheprocessofreplenishing an eroded beachwith new materialin order toservethesame purpose asgroynes. Therefore, it is an artificial process implemented by a human agent. In our knowledge base, it thus belongs to theADDITION category(artificialprocess),in the samewayas all other soft coastal defence techniques. Groyne as well as beach nourishment are members of the same mini-conceptualnetwork, which is more meaningful when placed within the context of the prototypicalcoastal engineering event. In spite of belongingto two different macro-categories (PROCESSandRESULT), the link between soft and hard beach stabilization approaches is evident since they share the samebasic function, and areboth types of coastal defence solutions. There are, of course, different ways of defining concepts, dependingon the objectivesofthe dictionary project and the lexicographers involved. However, we have opted for a definitional format that: (1) makes category membership explicit; (2) reflects aconcept’s relations with other concepts within the same category; (3) specifies essential attributes and features. As shall be seen, the definitions of groyne and beach nourishment fulfilthese conditions. The definitions that we haveformulated for coastal engineering terms arebased onthose found in other dictionaries and terminographic products that reflect the sharedknowledge of specialized concepts. The analysis of such resources makes it possible toextract basic meaning components, identify the type of component, and elaborate Linking Images and Words: the Description of Specialized Concepts definitions with external coherence on the basis of the same schema. For this type of analysis, three or more terminographic resources are used inorder toobtain a consensusof themeaning components for eachterm. As shall be seen, thestructure ofdefinitions ineach conceptual area has the same underlying template. Figure 2 is a segment of the definitional hierarchy in whichgroyne and beach nourishment are located. COASTAL DEFENCE Hard coastal defence structurempermeable coastal defence structure of concrete, wood, steel, and/or rock, which reflects a high proportion of incidentwave energy. hard coastal defence structure, perpendicularto the shoreline, whose function littoral drift,reducelongshore transport, and Breakwaterhard coastal defence structure, whose function is toprotect a shore area, harbour, anchorage, or basin from waves and weather. Detached Breakwaterbreakwater without any coastal connection to the shore. Soft coastal defence action: coastal defence action which aims to protect the coastline by enhancing natural processes and using natural forms of defence. Dune restoration: soft coastal defence action consisting of rebuilding an eroded dunethrough various processes like sand replenishment or revegetation. Beach nourishmentsoft coastal defence action consisting of replenishing a beach with from erosion, floods Figure 2. Segment of the definitional hierarchy of coastal defence. As shown in Figure 2, the genus or nuclear part the definitions is indicative of the IS_Arelationship,and as a result, marks membership in a specific category as well as degreeof specificity.In the case ofgroyne, the genus (i.e. hard coastal defence structure)shows:(1)its membership in the generic category of COASTAL DEFENCE; (2) its relationship to the immediately superordinate concept hard coastal defence structure; (3) its coordinate relationship with breakwater. As for beach nourishment, the genus (i.e. soft coastal defenceaction) highlights: (1) its membershipin the generic category coastal defence;(2)its relationship to the immediately superordinate concept soft coastal defence action; (3) its coordinate relationship with dune restoration. Furthermore, the differentiae or differentiating features specified in the definitions also provide valuable information because these are the characteristics that distinguish one conceptfrom anotherwithin the samecategory. In thissense, Barsalou (1992) affirms that features implicitly show attributes and values that should become explicit in frames. InTerminology, they also establish parameters for the creation and organization of more specific terminological units thatenable a multidimensionalrepresentationofconceptsinthe domain. The analysisofthe characteristics lexicalized in differentiae provides the defining featuresof groyne and beach nourishment, and also points to non-hierarchical relations that Linking Images and Words: the Description of Specialized Concepts must be taken into account in any multidimensional representationofthe concept. In fact, they constitute the parameters that categorize the various subtypes of groyne. These parameters can be observed in the template (Fig. 3)that underlies the definitions ofall the terms designating coastal defence structures. Figure 3. Activation of the definitional template for hard coastal defence structures in the definition of groyne. As shown in Figure 3, the definitional template forhard coastal defence structure has four types of conceptual relations: (1) IS_A; (2) MADE_OF; (3) HAS_LOCATION; (4)HAS_FUNCTION. In Fig. 4 the definitional template for soft coastal defence action also displays four types of conceptual relations: (1) IS_A; (2) RESULT_OF; (3) HAS_INSTRUMENT; (4) HAS_FUNCTION.Thus, a coastal defence template would activate the common relations IS_A and HAS_FUNCTION. In contrast, in the case of hard and soft solutions, conceptual relations vary in accordance with the nature of the concepts. Since hard coastal defence structures are objects andsoft coastal defence actions are regardedas processes, theirdefinitions showdifferent types ofinformation: theMADE_OFconceptual relation pertains to the description of aphysical entity, whereas the RESULT_OFconceptual relation focuses on the final phase of the process involved. Generally speaking, the linguistic description of groyne andbeach nourishment aswellas all of their coordinate and subordinate concepts within the conceptual area tend to follow these templates insofar as type, quantity, and configuration of information are concerned. In this way, definitions show a more uniform structure that complement the information encoded in the conceptual system, and directly refer to and evoke the underlying event structure of the domain. These templates can even be considered a conceptual grammar for the description of all types of coastaldefencewithin the Coastal Linking Images and Words: the Description of Specialized Concepts Engineeringdomain.This ensures a highdegree ofsystematisation at the micro-structural level. Figure 4. Activation of the definitional template for soft coastal defence actions in the definition of beach nourishment. On amacro-structural level, the templateisconsistent withthe dimensions specified for the designation of both concepts as shown in the network inFig. 5. Non-hierarchical relations are thus essential for dynamic knowledge representationsbecause they enhance conceptual structure by enriching networksand encoding multiple relations between concepts (Faber 1999). 2.1.2. Dredge. Our approach can also be applied to processes like dredge, whose linguistic designation is a verb form. Though linguistically, both groyne and beach nourishment are noun phrases, conceptually speaking, groyne is an objectand beach nourishment is a process. However, up to the present, verbs have receivedrelatively littleattention in specialized lexicography5 though they designate a myriad of states and processes that are instrumental in the representation of specialized domains. In the coastal engineering domain, dredge represents the action of removing sand inorderto renourish abeach.As showninpreviousexamples, thisactivity is also carried out by ahuman agent. Consequently, dredge belongs tothe SUBTRACTION category (artificial process). This shows how the CEE can be extremely useful in the categorization of this particular domain. The three examples in this paper belong to different subcategories which are, notwithstanding, interrelated at the macro-structurallevel (CEE, Fig. 1)aswell Linking Images and Words: the Description of Specialized Concepts as at the micro-structural level, wheredifferent categories oftenconverge in the same sub-hierarchy (Fig. 5). Figure 5. Conceptual network for groyne andbeach nourishment. Dredge is closely linked to beach nourishment sinceitobviouslyrepresents theactionof dredging or the means of extracting sand to replenish an eroded beach. It consists ofa sequenceofsubprocesses. This dynamism is not only reflected in the state of affairs oftheverb, butalsoin its arguments, whichbelong todifferent categories within the event. (e.g. Beach can beaPATIENT aswellas a RESULT). Thisis yet another argumentinfavourofaflexible, dynamic, and multidimensional process-oriented event as the underlying structure forthis knowledge area. The elements included in the definition of dredge are represented in the conceptual networkinFig.5, and are basicallythe sameas dredging. Since dredge is a process,its representation includes the action itself, the phases of the action, and the instrumentsused tocarryitout.Inthis case,beach nourishment canbe regardedas thefunctionor purpose of dredge. The identification of category membership for verbs requires adifferent treatment. In concepts represented by verbs (i.e. actions or processes), the IS_Arelationship, while still marking the degree of specificity, is specifically based on the participantsin the event (argumentstructure) and the conceptual category to whichthe concept belongs. The theoretical priniciples of the Functional Lexematic Model (FLM)(Martín Mingorance 1984, 1989, 1995;Faber and Mairal 1999)can be used for the conceptual organization of verbs. Thismodelproposes alexicon withtwoaxes:aparadigmatic axis, which organizeslexemes hierarchically, and a syntagmatic axis, which describes the complementation patterns of each verb (Faber and Mairal 1999). Through the lexical decomposition of Linking Images and Words: the Description of Specialized Concepts lexicographic definitions,verb concepts are placedin different domains andsub-domains within the paradigmatic axis. The verb lexicon is thus divided into twelve basic lexicaldomains, each of which can be said to have its own definitional template. The nuclear meaning ofeach verbis representedby asetofnear-primitives, and lexicaldomains are divided into subdomains in accordance with semantic parameters or differentiae. The inventory of lexical domains is shown in Figure 6. Existence Change Possession Speech Emotion Action Manipulation Cognition/mental perception Movement Generalperception Sense perception Position Figure 6. Lexical domains in the FLM. Based on the above list, dredge belongs tothegeneralcategory ofmovementinthesame way as other similar action concepts in the coastal engineering domain, such as drain, pipe or excavate. However, not all of them necessarily have the same underlyingstructure. movement verbs are defined in terms of generic terms such asmove,go orcome.However, assemantic parameters are added to the paradigmthrough the designation oftypical arguments, the hierarchy becomes more specific because the arguments (or participantsin the process) are coastal engineering concepts. Thus,verb concepts sharing thesame definitionaltemplateare those thathavethe samegenus and thesameargumentstructure as shown in Fig. 7. MOVEMENT To cause somebody/something to move away from a place/thing/person remove [arg1,AGENT]tomove away somebody/something [arg2, PATIENT] from a place [arg 3,LOCATION] dredge 1 [arg1,AGENT]toremove material [arg2, PATIENT] fromthe bottom of any body of water. [arg 3,LOCATION] Linking Images and Words: the Description of Specialized Concepts dredge 2 [arg1, AGENT]to remove beach material [arg2, PATIENT]from underwater [arg 3, LOCATION]by pumping,extracting and piping it by means of a dredger for beach nourishment. drain [arg1, AGENT] to remove liquid [arg 2, PATIENT] from any surface [arg 3, LOCATION] Figure 7. Organization of verb concepts in the paradigmatic axis. In this case, the genus of dredge (remove)marks its membership inMOVEMENT,and more specifically, in the subdomain of MOVEMENT_AWAY, whose members show a similardefinitional template. For instance, acoherent and exhaustive terminographic description of drain would also include thesesame elements, whereas MOVEMENT verbs belonging to other subdomains would possibly show other patterns. Remove, the genus in the definitions of MOVEMENT_AWAYverbs, can imply that someone (arg 1) moves something (arg 2) away from somewhere (arg 3). These arguments (AGENT, PATIENT and LOCATION)function asdefault values within the same paradigm distinguishing coordinate concepts. At the same time, depending on the specificity of its arguments, a concept can be regarded as moreor less specialized, and thus beframed in a more restricted domain. The definition of dredge1, extracted fromageneral language dictionary, describes avery general event, whereas dredge2 unequivocally refers to the coastal engineering domain, and more precisely to the beach nourishment process. The definitional template for verbs withinthis paradigm is shownin Fig. 8. Figure 8. Activation of the definitional template of MOVEMENT_AWAYverbs in the definition of dredge. In Fig. 8 the definitional template of MOVEMENT_AWAY concepts displays three types of conceptual relations: (1) IS_A;(2)HAS_PATIENT;(3)HAS_LOCATION. However, the specificity Linking Images and Words: the Description of Specialized Concepts of dredge lies in the selection restrictions for the arguments, namelythefact thatwhatisremoved is an object not aperson, andmore specifically, beach material. Furthermore,thelocation is restricted to underwater. The representation of dredge also includes three more conceptual relations,whichdepict how thismovementiscarried out, the instrument used, and the reason why it is done. In Fig. 8 the definitional template of MOVEMENT_AWAY concepts displays three types of conceptual relations: (1) IS_A;(2)HAS_PATIENT;(3)HAS_LOCATION. However, the specificity of dredge lies in the selection restrictions for the arguments, namelythefact thatwhatisremoved is an object not aperson, andmore specifically, beach material. Furthermore,thelocation is restricted to underwater. The representation of dredge also includes three more conceptual relations,whichdepict how thismovementiscarried out, the instrument used, and the reason why it is done. 2.2 Typology of graphical information However definitions are not the only means of describing concepts. Images are also an important resource for this purpose. The inclusion of different types of visualrepresentation is extremely helpful in specializedknowledge fields sinceimagesenhance textual comprehension and complement the linguistic information providedin other data fields. Engineering dictionaries generally include occasional illustrations as a way of offering a more completedescriptionof certain concepts (e.g. Flack and Möllerke 1999; Ching and Adams2000; SmitandChandler 2000; Adrien2003).It isalso worth noting that the Le grand dictionnaire terminologique (http://www.granddictionnaire.com), oneofthe most importantterminologicalresources inCanada also contains illustrations as part ofsome of the term entries. In addition, almost all terminology management tools available onthe market today provide the option for including graphics (e.g. MultiTerm, Termbase, etc.). Even thoughmost terminologists and lexicographers do recognize the value ofgraphicalinformation and make useofit whenever possible,illustrations are sometimesinsertedina relatively haphazard way in lexicographical and terminological resources without sufficient reflection on the visual representation of concepts. In our opinion, graphical material should be chosen in a principled way so as to be consistent with linguisticdescription. Traditionally, images have beenclassified according to their morphology in categories ofphotographs, drawings, animations, videos, diagrams, charts, graphics, schemes, views, etc. (Darian 2001; Monterde 2002). However, it is more useful to categorize images in terms of theirmost salient functions (Anglin et al. 2004)or intermsof their relationshipwith the real-world entitythat theyrepresent. Wehave thusbased our typology ofimageson the criteria of iconicity, abstraction and dynamism as ways of referring to and representing specific attributes of specialized concepts. . Iconic images resemble the real-world object represented through the abstraction of conceptual attributes in the illustration. Linking Images and Words: the Description of Specialized Concepts . Abstraction in an illustration is amatter of degree, and refersto the cognitive effort required for the recognition and representation of the concept thus represented (Levie and Lentz 1982; Park and Hopkins 1993; Rieber 1994). . Dynamismimplies the representationofmovement(i.e. video andanimation, as well as images showing different stages of a superordinate process respectively).However, such arepresentation need not include explicit movement if it illustrates the succession of discrete steps that make up theprocess. 3 Linking conceptual descriptions and images 3.1 Groyne and beach nourishment As previously mentioned, the inclusion of images in term entries is based on the conceptual relations activated in the definition of the concept. In this respect, and in accordance with the definitional template for hard coastal defence structure and soft coastal defence action, the graphical resources inour image database highlight at least one ofthe following conceptual relations, IS_A, HAS_LOCATION, MADE_OF, HAS_INSTRUMENT_OF,RESULT_OFand HAS_FUNCTION,for the concept that they represent.They achieve this throughiconicity, abstraction, and/or dynamism. Iconicityiswidely usedin the representationofdifferent types ofcoastal structures.The resemblance of the illustration to the real-world entity allows the user to identify the object or process in question by inferring its basic characteristics and linking them to previously stored knowledge structures. Iconicity is present inthe linguistic description of the conceptas well sincethe characteristics of thegeneric term areincludedin thegenus of the subordinates. The use ofa definitional template means thatallof the definitions ofcategory members will resemble each otherto some degree. Iconic images are especially prevalent when it isnecessary to represent non-hierarchical relations, such as MADE_OF,whichlink the structure tothematerial thatitiscomposedof (Fig. 9), andHAS_INSTRUMENT which shows the entity used to carry outan action (Fig. 10). Figure 9. Concrete groyne: representation of the conceptual relationMADE_OF. Linking Images and Words: the Description of Specialized Concepts Thelevel of abstraction can facilitatetheunderstanding ofconceptual relations such as HAS_LOCATION andIS_A. For example, in regards to groyne, location can beunderstood as the position of the coastal structure in reference to the shoreline (Fig. 11), or it can be understood as its position in referenceto the sea bottom (Fig. 12). Theabstraction of these images highlights this particular relation, while others such as groyne type, material, or function recede into thebackground and are not focused on. Figures 11 and 12. Representation of the conceptual relation HAS_LOCATION[POSITION]andHAS_LOCATION[SPATIAL CONFIGURATION]. On the other hand, since beach nourishment is a process, it can be understood as a sequence of phases. In this case the IS_A relation is encoded by images with a certaindegree of abstraction. As shown in Fig. 13, the use of complementary short textual descriptionsfacilitatestheunderstanding ofits temporal and spatial location aswell asall of thefactors of the process. Linking Images and Words: the Description of Specialized Concepts Accordingly, since beach nourishment is an action carried out over a period oftime,the relationRESULT_OF is best represented by abstract images that makeexplicit reference to the temporal nature of the process (Fig.14). Linking Images and Words: the Description of Specialized Concepts Finally, the relation HAS_FUNCTIONis generally represented by dynamic images. Functionis one of the most important features for the description of both hard and soft coastaldefence solutions since such structures are often constructed or implemented to cause acertain (positive) effect, but all too often bring about negative changes in the coastal environment that had not been foreseen. Since the illustration of a function means representing a process, dynamism is usually present, though evidently, this does not preclude the presence of degrees of iconicity or abstraction. Figs. 15a and 15b are twodynamic illustrations that represent the positive and negative effects of groynes on a shoreline. Figures 15a and 15b. Groyne-adjusted shoreline: representation of the conceptual relation HAS_FUNCTION. Sometimes one illustrationmay not be enough to represent complex conceptual relationssuch as HAS_FUNCTION. The combination of two dynamic images (Figs. 16a and 16b)implements the protective function stated in the definition. Fig. 16a shows the different coastal processes responsible forsand gains and losses, whereas Fig. 16b represents the sand balance, whose restorationis the primary objective ofbeach nourishment. Figures 16a and 16b. Combination of two dynamic images: representation of the conceptual relation HAS_FUNCTION. In these figures, dynamism is conferred by the use of symbols, such as arrows (representing movement), and textual informationthat link thepictures to the real world. Nevertheless, since they are non-iconic elements, a certain level of abstraction is also involved in arriving at the understanding that groynes retard longshore transport and littoral drift at the same time as beach nourishment provides a deposition of dredgedsediments. As a consequence, they jointly contribute to an accretion of sediments in order to prevent beach erosion. Linking Images and Words: the Description of Specialized Concepts Consequently, the entries for groyne andbeach nourishment are accompaniedby aset of images whose iconicity, abstraction, and/or dynamism are consistent with the information in their definition or linguistic description. The visualfocus ofeachimageiscentred on a specific conceptual relation that is lexicalized in the terminographic definition. The linking of linguistic and graphical information in accordance with a definitional template typical of the conceptual area affords a more complete description and a better comprehension of specialized concepts (Figs. 17and 18). 3.2 Dredge Given the fact that dredge is a process, theIS_Arelation is best conveyed by means of adynamic image such as Fig. 19 which depicts the sequence of sand removal, pumping, piping and deposition. The combination of explicative text boxes and non-iconic elements, such as arrows, reduces the level of abstraction and facilitates the understanding of this concept. Linking Images and Words: the Description of Specialized Concepts Dynamic images, which are somewhat less abstract, are also suitable for the representation of the RESULT_OF relation, which is the final phase of the dredge process.Thecombinationof illustrations inFig. 20provides the userwith an overall view of the stages in this process: pumping, extracting and piping. These process phases can be represented from adual perspective, an aerial view and also ahorizontal cross-section, which shows how the sand goesfrom the sea bottom onto thebeach. Linking Images and Words: the Description of Specialized Concepts Another important conceptual relation for dredge is HAS_LOCATION, which represents where the dredgingequipmentis situated. Since thisrelationis very easy tograsp, Fig. 21 hasaverylowlevelofabstraction, and merelyshowstheunderwater location where the dredging process starts. The caption as well as iconic elements (e.g. darkandlightshades for the representation ofthe sea background andthe sea bottom asreal-world entities)are also present in the illustration itself and contribute to the representation of HAS_LOCATION. The conceptual relation HAS_FUNCTIONonce again reflects the procedural nature ofdredge.Consequently, dynamic illustrations depict the phases of the process. Fig. 22 shows the beachfill (orbeach nourishment), which is the primary objective ofa dredging project. Linking Images and Words: the Description of Specialized Concepts Iconicity is particularly significant when representing the conceptual relation HAS_INSTRUMENT since the different types of dredgers (hydraulic and mechanical) differ from one another in the position of the pumps and pipes. Iconic images reflecting the possible positions ofthe articulated piping equipment contribute to the representation of HAS_INSTRUMENT, becausethey greatlyresemble the actual dredger vessel and the position of its dredging equipment (Figs 23, 24and 25). Figures 23, 24 and 25. Different types of dredgers: representation of the conceptual relation INSTRUMENT_OF. Linking Images and Words: the Description of Specialized Concepts In the same way, the conceptual relation HAS_PATIENT is iconically represented by a photograph of beach material as it is being dredged. Its resemblance to the real-world entity facilitates the identification of the material removed (sand or shingle) when dredging. The combination of both iconicity and dynamism highlights the role of beach materialas thepatient ofdredge. The three parameters used to classify illustrationsin order toenable the representation ofthe conceptual relations behave rather differently when applied to procedural concepts designated by verbs. There is certainly moredynamicity in theimages that represent such concepts, since most conceptualrelations, IS_A, RESULT_OF and HAS_FUNCTION highlight the stages of the process in question. Similarly, these illustrations show a low level of abstraction which reduces complexity normally present in procedural concepts (Fig. 27). 4. Conclusions In this article we have underlined the importanceofinterrelatingdifferent types ofdataina terminological knowledge basein order to achieve ahigh degree of internal and external coherence. This is a complex issue since a database that aspires to be a knowledge resource cannot be a mere repository for bits and pieces of informationthat have been scavenged from different sources and gathered together in a kind of terminographic magpie’s nest. Our assertionis thatallelementsinthedatafields of term entries must be interrelated. The entries for groyne, beach nourishment, anddredge reflect their location within the Coastal Engineering Event. The linguistic descriptions of these concepts are consistent with the definitions ofothersinthesamearea. The conceptual relationslexicalizedinthecategory templates abstracted from definitional structure form a relational set that ischaracteristic of other related coastal structure entities. Linking Images and Words: the Description of Specialized Concepts In aparallel way, the graphical description of concepts shouldnot be random. Such imagescanbe chosen in a principledwayso asto be in accordance with linguistic description. Illustrationsshouldbe selectedso thatthey(1)focus ononeor moreaspectsof the core templateinformation as activated in the linguistic description oftheconcept; (2) possess the level of iconicity, abstraction, and/or dynamismthat best portrays the attribute/s of the concept. Linking Images and Words: the Description of Specialized Concepts In conclusion, linguistic and graphical descriptions ofspecialized entities play amajor role in the understanding process when both types of description converge to highlight the multidimensional natureofconcepts (Bowker 1997; Bowker and Meyer 1993) as well asthe set of conceptual relations typical of a specific subdomain. Notes 1. This article isarevisedandexpanded versionof a paper presentedat the2006 Euralex Conference in Turin, Italy. The research is part oftheproject PuertoTerm: knowledge representation and the generation of terminological resources within the domain of Coastal Engineering, BFF2003-04720,fundedby the Spanish MinistryofEducation. 2. The authors would like to thank the two anonymous reviewers for their helpfulsuggestions. 3. Bowker and Meyer (1993) and Bowker (1997) advocate a multidimensional representation for concepts in specialized domains. 4. PuertoTerm is a research project whose principal objective is the generation ofterminological resources for the representation of specialized concepts in the domain of Coastal Engineering. 5. An exception to the general tendency in lexicography and specialized lexicography to focus on nouns and ignore verbs is the work done by Fellbaum (1990, 1998) and L’Homme (2002, 2003, 2006). References Adrien, Nicholas 2003. Computational Hydraulics and Hydrology: An Illustrated Dictionary. Boca Raton, FLA: CRC Press. Anglin, G. Vaez, H. and Cunningham, K. 2004. ‘Visual Representations and Learning: the Role of Static and Animated Graphics’ in Jonassen, D. (ed.) Handbook of Research on Educational Communications and Technology. Hillsdale, NJ: Lawrence Erlbaum, 755 794. Atkins, Sue, Rundell, Michael, and Hiroaki, Sato 2003. ‘The Contribution of Framenet to Practical Lexicography.’International Journal of Lexicography 16(3): 334-357 Barsalou, Lawrence W. 1992.‘Frames,Concepts,andConceptualFields’inLehrer,A.and Kittay, E.F. (eds.),Frames, Fields and Contrasts: New Essays in Semantic and Lexical Organization. Hillsdale, NJ: Lawrence Erlbaum, 21-71. Barsalou, Lawrence W. 2003. ‘Situated Simulation in the Human Conceptual System.’ Language and Cognitive Processes 18, 513-562. Béjoint, Henri 1997.‘Régards sur la définition en terminologie.’ Cahiers de Lexicologie 70 (1): 19-26. Boas, H. C. 2005. ‘Semantic Frames as Interlingual Representations for MultilingualLexical Databases.’ International Journal of Lexicography 18(4): 445-478. Bowker, Lynne 1997.‘MultidimensionalClassification of Concepts and Terms’ in Wright, S.E. and G. Budin (eds.), Handbook of Terminology Management.Philadelphia/Amsterdam: John Benjamins, 131-143. Bowker, Lynne 2003. ‘Specialized Lexicography and Specialized Dictionaries’ inSterkenburg, P. van (ed.), A Practical Guide to Lexicography.Philadelphia/Amsterdam: John Benjamins, 154-164. Linking Images and Words: the Description of Specialized Concepts Bowker, Lynne and Meyer, Ingrid 1993.‘Beyond‘Textbook’ConceptSystems: Handling Multidimensionality in a New Generation of Term Banks’ in K.D. Schmitz (ed.), TKE'93: Terminology and Knowledge Engineering, Frankfurt:Indeks Verlag, 123-137. Ching, Francis D. K. and Adams, Cassandra 2000. Building Construction Illustrated. New York: Wiley. Darian, Steven 2001.‘More thanMeetstheEye: the RoleofVisualsinScience Textbooks.’ LSP & Professional Communication 1 (1): 10-36. Faber, Pamela 1999. ‘Conceptual Analysis and Knowledge Acquisition in ScientificTranslation.’ Terminologie et Traduction 2:97-123. Faber, Pamela 2002. ‘Terminographic Definition and Concept Representation’ in Maia, Belinda, Haller, Johann, and Ulyrch, Margherita (eds.), Training the Language Services Provider for the New Millennium. Oporto: University of Oporto, 343-354. Faber, Pamela and Mairal Usón, Ricardo 1999. Constructing a Lexicon of English Verbs. Berlin: Mouton de Gruyter. Faber, Pamela, López Rodríguez, Clara, and Tercedor Sánchez, María Isabel 2001. ‘Utilización de técnicas de corpus en la representación del conocimiento médico.’ Terminology 7:2, 167-197. Faber, Pamela, Márquez Linares, Carlos, and Vega Expósito, Miguel 2005. ‘Framing Terminology: A Process-Oriented Approach’. META 50 (4):CD-ROM. Faber, Pamela, Montero Martínez, Silvia, Castro Prieto, María Rosa, Senso Ruiz, José, Prieto Velasco, Juan Antonio, León Arauz, Pilar, Márquez Linares, Carlos, and Vega Expósito, Miguel (2006), ‘Process-oriented Terminology Management inthe Domain of CoastalEngineering.’Terminology 12:2, 189-213. Fellbaum, Christiane. 1990. ‘English Verbs as a Semantic Net’. International Journal of Lexicography 3 (4), 279-301. Fellbaum, Christiane.(1998). ‘Introduction’ in C.Fellbaum (ed.) Wordnet: An Electronic Lexical Database. Cambridge, MA:MIT Press, 1-20. Fillmore, Charles J. 1976. ‘Frame Semantics and the Nature of Language’ in Annals of the New York Academy of Sciences: Conference on the Origin and Development of Language and Speech 280, 20-32. Fillmore, Charles J. 1982. ‘Frame Semantics’ in Linguistics in the Morning Calm, Seoul: Hanshin Publishing Co., 111-137. Fillmore Charles J. 1985. ‘Frames and the Semantics of Understanding.’ Quaderni di Semantica 6:222-253. Fillmore, Charles and Atkins, Sue 1998. ‘FrameNet and Lexicographic Relevance’ in Proceedings of the ELRA Conference on Linguistic Resources, Granada, 417-423. Fillmore, Charles J., Petruck, Miriam R.L., Ruppenhofer, Josef, and Wright, Abby2003. ‘Framenet in Action: the Case of Attaching.’ International Journal of Lexicography 16 (3): 298-332. Flack, H. K. and Möllerke, G. 1999. Illustrated Engineering Dictionary: Bildwörterbuch Maschinenbau und Elektrotechnik Englisch/Deutsch Deutsch/Englisch (VDIBuch).Berlin: Springer. Grinev, S. and Klepalchenko, I. A. 1999.‘Terminological Approachto KnowledgeRepresentation’ in Sandrini, P. (ed.), '99. Terminology and Knowledge Engineering, Vienna: TermNet, 147-151. Linking Images and Words: the Description of Specialized Concepts Levie, W. H. and Lentz, R. 1982. ‘Effects of Text Illustrations: a Review of Research.’ Educational Communication and Technology Journal 30: 195-232. L’Homme, M. C. 2002. ‘What can Verbs and Adjectives Tell us about Terms?’ in Terminology and Knowledge Engineering, TKE 2002. Proceedings, August 28-30, 2002, Nancy, France, 65-70. L’Homme, M. C. 2003. ‘Capturing the LexicalStructure in Special Subject Fields with Verbsand Verbal Derivatives: A Model for Specialized Lexicography.’ International Journal of Lexicography 16(4):403-422. L’Homme, M.C. 2006. ‘Using Explanatory and Combinatorial Lexicology to Describe Terms’in Wanner, L. (ed.).Selected Lexical and Grammatical Topics in the MeaningText Theory. In Honour of Igor Mel'cuk, Amsterdam/Philadelphia: John Benjamins. Martin, W. 1998. ‘Frames as Definition Models for Terms’ in Proceedings of the International Conference on Professional Communication and Knowledge Transfer.Vienna:Termnet, 2:189-221. Martin, W. (2006).‘Frame-based Lexicons and the MakingofDictionaries.’Proceedings of the 12th Euralex International Conference, Turin, Italy. Martín Mingorance, L. 1984. ‘Lexical Fields and Stepwise Lexical Decomposition in a Contrastive English-Spanish Verb Valency Dictionary’ in Hartmann, R. (ed.). LEXeter 83: Proceedings of the International Conference on Lexicography. Tübingen, Germany: Max Niemeyer, 226-236 Martín Mingorance, L. 1989. ‘Functional Grammar and Lexematics’ in Tomaszcyk, J. and Lewandowska, B. (eds.). Meaning and Lexicography, Amsterdam/Philadelphia: John Benjamins, 227-253. Martín Mingorance, L. 1995. ‘Lexical Logic and Structural Semantics: Methodological Underpinnings in the Structuring of a Lexical Database for a Natural Language Processing’ in Hoinkes, U. (ed.). Panorama der Lexikalischen Semantik, Tübingen, Germany: Gunter Narr, 461-474. Monterde Rey, A. M. 2002. ‘Terminología: estudio de las distintas formas derepresentación conceptual en textos técnicos y su relación con la traducción’ inActas de las II Jornadas de Jóvenes Traductores. Las Palmas de G. C.: Servicio de Publicaciones de la Universidad de Las Palmas de G. C., 147-156. Park, O. and Hopkins, R. 1993. ‘Instructional Conditions for Using Animated Visual Displays: AReview’. Instructional Science 22:1–24. Rieber, L.P. 1994. Computers, Graphics, and Learning. Madison, WI: Brown & Benchmark. Smit, Kornelis and Chandler, Howard 2000. Means Illustrated Construction Dictionary. (First edition 1991) Kingston, MA: R. S. Means Construction Publishers and Consultants.