2.2.1 Choosing which relations to study
There are many interesting semantic relations that provide avenues for study: hyponymy, meronymy, functionality, synonymy, antonymy, causality, etc. However, the present thesis is limited to the first three.
Given that the secondary purpose of this thesis is to contribute to the development of the Text Analyzer (TA), I limited myself to working with, at most, the four relations that the TA can already handle: hyponymy, meronymy, functionality, and synonymy. I later decided not to study synonymy after all because the source (TERMIUM) I used to discover the linguistic patterns for the chosen relations turned out to be poor in explicitly stated synonymous relationships.
While it may at first seem that studying only three relations is not very extensive, at second glance, we realize that each relation is complex, as described in the following sections.
2.2.2 Hyponymy
Hyponymy is a relation of inclusion. In the words of Lyons (1968, 453), "...the meaning of tulip [or rose or geranium] is said to be 'included' in the meaning of flower." To take another example from Lyons (1968, 453), "the 'meaning' of scarlet is said to be 'included' in the 'meaning' of red." When represented hierarchically, "The 'upper' term is the superordinate and the 'lower' term the hyponym" (Palmer: 1981, 85).
| FLOWER | ||
|---|---|---|
| TULIP | ROSE | GERANIUM |
Therefore, the term tulip is a hyponym with respect to flower, while tulip, rose and geranium are co-hyponyms, i.e. hyponyms of the same superordinate term. Looking at this from the opposite angle, flower is the superordinate with respect to tulip, rose and geranium. (It should be noted that the designation of a term as a superordinate or hyponym is relative, not absolute. For example, tulip is hyponym of flower, but in turn would be a superordinate of parrot tulip).
This relation is also called generic-specific or taxonomy(4). Lyons (1977, 291) defines hyponymy as "the relation which holds between a more specific, or subordinate, lexeme and a more general, or superordinate, lexeme." Lyons (1968, 453) explains that "This relationship...has been formalized by certain semanticists in terms of the logic of classes". Indeed, "From the time of Aristotle this relation [taxonomy] has been central to the process of definition. The classical form...starts with a genus which is...superordinate to the term to be defined. To this are added the differentiae which distinguish this term from related terms" (Evens et al: 1980, 119). The "related terms" just mentioned are co-hyponyms.
Regarding the specific linguistic structures that express hyponymy, semanticists and other scholars have indicated some as part of their discussion on semantic relations. Cruse claims that "A useful diagnostic frame for taxonomy is: An X is a kind/type of Y" (1986, 137). The X and Y are "slots" that represent any logically and grammatically correct terms such as "A spaniel is a kind of dog". Cruse raises an issue that is relevant to the problem of noise (i.e. sentences that contain this structure but do not express the relation in question). He states that "the expression kind of is not univocal, and it is necessary to be able to recognize those senses which are irrelevant for the diagnosis of taxonomy" (1986, 137-8). For example, "he was wearing a kind of flattened, three-sided turban--I don't know exactly what it was" (1986, 138). This statement is simply an approximate description of an unfamiliar item and should not be mistaken as hyponymy. Cruse also proposes diagnostic frames for co-taxonymy (co-hyponymy): "An X is a kind of Y, and a Z is another kind of Y" and "...and so is a Z". The expressions another kind of and so is both imply that X is not the only member of the class Y, that Z, too, belongs in this class. Cruse also suggests a frame for verbs, which, as he says, "seem to show hierarchical structuring to a more limited extent than nouns" (1986, 139). For example, he mentions "X-ing is a way of Y-ing", as in "walking is a kind of moving".
Lyons, too, has determined some of the linguistic structures indicating hyponymy. "When...hyponymy holds between nouns, it is possible to insert syntactically appropriate expressions containing them in place of X and Y in the following formula 'X is a kind of Y'... 'sort' and 'type' may be substituted for 'kind' in colloquial English" (1977, 292). He also suggests that "There are many other more specific lexemes...which may be employed...eg. 'shade' in 'Crimson is a shade of red" (1977, 292-3). Indeed, this is a hint that, within subject fields, there may be structures that are particular to that field, as discussed later in this thesis. Regarding co-hyponymy, Lyons says that "When a noun X is superordinate to more than one hyponym...such expressions as the following will be accepted as meaningful: 'cows and other (kinds of) animals'" (1977, 293). This implies that cows are not the only member in the class of animals, which is obvious to most people, but for terminologists working in an unfamiliar field, this type of implied knowledge is very informative by indicating that there are further hyponyms to watch for.
Ahmad and Fulford (1992, 13) prepared a list of such potential structures for hyponymy (and four other relations) and tested some of them for effectiveness on a corpus. The list includes form%, type of, is a, and group* (5). Pearson, who dealt with hyponymy only, as expressed in formal definitions, discovered different "hinges" (1996, 820) depending on whether X comes before the defining statement or after. Some examples of hinges are: X is/are, X consist(s) of, X is/are defined as, and is called a X. Pearson (1996, 821) expresses concern about focussing adverbs such as generally and usually because they "prevent statements from having generic reference....When a term is described as usually having a particular characteristic, it is not possible to conclude that it always has this characteristic". I believe that excluding sentences containing these adverbs is fine under ideal conditions. However, the fact that "X is usually blue" still constitutes knowledge, and as such is valuable information for terminologists, who are interested in the actual state of knowledge as it exists in a subject field, and not in ideal expressions of relations.
Borillo (1996, 113) writes that "En linguistique, on s'intéresse en général à la relation d'hyperonymie dans le cadre de la construction de dictionnaires, pour l'élaboration de la définition des mots". She conducted a study, the objective of which was to "voir comment la relation d'hyperonymie se manifeste linguistiquement dans les textes, i.e. sous quelles formes lexico-syntaxiques elle trouve à s'exprimer" (Borillo: 1996, 114) and later "repérer dans les textes des structures très locales susceptibles d'être interprétées comme des mises en relation de type hyperonymique" (Borillo: 1996, 121). The study revealed a wide variety of French structures, including (but not limited to) the following:
Hearst (1992) conducted a similar study for English and started with the following list of patterns: such...as, or other, and other, including, and especially. The Hearst study was also concerned with describing a method for automatic discovery of new patterns.
2.2.3 Meronymy
In the words of Cruse (1986, 157), "there is no doubt about the central importance of fully integrated and cohesive physical objects, with well-differentiated parts, in the concepts of 'part' and 'whole'". Van Campenhoudt (1996, 60) describes the nature of meronymy as follows: "Le principal élément relationnel est la connexion du tout et de ses parties". He also explains the importance of meronymy for knowledge-based terminology: "Une base de connaissances terminologiques devrait notamment être à même de gérer les relations méronymiques à travers différentes tâches" (1996, 79).
Cruse (1986, 158) provides an amusing description of the difference between "piece" and "part": "The contrast between parts and pieces is potentially operative even with highly integrated wholes such as animal bodies: there is a clear difference between such a body hacked to pieces, and one carefully dissected into its parts". It is important to note at this point that meronymy is not limited to physical objects. In the words of Cruse, "one may also speak of parts of non-concrete entities such as events, actions, processes, states, and abstract nominal notions" (1986, 172). An example of this is the following: "The steps in WordPerfect 7.0 for italicizing text are a) highlighting the desired text, b) clicking the italics button on the toolbar, and c) clearing the highlighting."
The following diagram shows how part-whole relationships can be represented hierarchically:
| CAR | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Body | Wheel | Electrical System | Etc. | ||||||||
| frame | chassis | etc. | tire | rim | etc. | starter motor | battery | etc. | etc. | etc. | etc. |
There are different ways in which something can be a part of something else, and this has led scholars to claim that meronymy is a complex relation. Iris et al say, "the part-whole relation should be treated as a collection of relations, not as a single relation" (1988, 262). On this point, Miller asserts that "Such observations raise questions about how many different 'part of' relations there are" (1990, 256). Iris et al (1988) describe "four models of the part-whole relation": the functional component (the most common); the segmented whole; collections and members; and sets and subsets. Winston et al (1987) elaborated a "taxonomy of part-whole relations" that involves six types of meronymy:
They also propose three "relation elements": functional, homeomerous, and separable, for which each meronymy type is either yes (+) or no (-). The relational element "functional" indicates that the part has a function with respect to its whole; "homeomerous" means that the part is identical to the other parts making up the whole; and "separable" means that the part can be separated from the whole. For example, the relation between handle/cup is "component/integral object" and is functional (+), homeomerous (-), and separable (+). This means that the handle has a specific function with respect to the whole, does not resemble the other parts of the cup, and can be separated from the whole. In a later study (Chaffin et al, 1988), four more relational elements are added.
Given that there is more than one way X can be part of Y, it stands to reason that there exists a variety of linguistic structures that can express meronymy. Winston et al focus only on part of and its relatives: The X is part of the Y; X is partly Y; X's are part of Y's; X is part of Y; the parts of a Y include Xs, etc, but they grant that "knowledge of parts and wholes can be expressed in many specialized ways" (1987, 417-418). They do propose such words as component, member, portion, and feature, all which can substitute for part in specific instances (1987, 430).
Cruse discusses test-frames for meronymy, such as "A Y has Xs/ an X" (1986, 160) and "X is a part of a Y" (1986, 161), but says either one on its own is not a reliable indicator of meronymy; both can be used to express non-meronymic relations, such as "changing diapers is part of being a mother" (although I would argue that the large task of motherhood could perhaps be broken down into smaller tasks (parts) such as changing diapers). He states that "The parts of Y include the X/Xs" is a "test-frame that does not leak" (1986, 161); however, this is an ideal structure that will not always occur in actual language in use.
In the French language, Larivière points out comprend and se compose de as structures that come under the category of composition (1996, 412).
Miller explains that three types of meronymy are coded into WordNet: Wm is a component part of Wh; Wm is a member of Wh; and Wm is the stuff that Wh is made from, and that the is a component of relation is the most frequent (1990, 256).
Evens et al have found that "In English the part-whole relation seems to be expressed most often with have, of, or the possessive" (1980, 188-9). Lyons (1977, 312) points out that earlier works have claimed the same thing: "part-whole relations between lexemes are bound up with a particular sub-class of possessive constructions, exemplified by such semantically...related phrases and sentences as 'John's right arm' and 'John has a right arm'".
The Ahmad and Fulford study (1992) tested the efficiency of 17 linguistic structures expressing what they refer to as the partitive relation. Their list includes consist* of, constituent%, factor%, zone%, contain*, and component% (1992, 15). Interesting to note is that Ahmad and Fulford see material as a relation separate from meronymy (1992, 13), whereas other scholars consider it to be a type of meronymy.
Jackiewicz (1996) conducted a study in French whose purpose was to "montrer par quels moyens lexicaux la langue individualise les parties d'un tout" (51). This study focussed on verbs and nouns and classified the findings into three main categories: words that express 1) action de composition d'objet, 2) action de décomposition d'objet, and 3) types des parties. The first category includes such words as réunir, joindre, ensemble, liaison, construction and agrégat. A sample sentence is "Le granite résulte de la fusion de roches détritiques" (Jackiewicz: 1996, 58). The second category includes désassembler, diviser, and fragmentation. An example given is "L'eau s'analyse en oxygène et hydrogène" (60). The third category includes composant, élément, and stade, and a sample sentence is "L'opium figure parmi les ingrédients essentiels de la Lamaline" (61).
The ultimate conclusion to be drawn from this discussion is that meronymy is a complex relation. In addition, there appears to be no consensus on either the types of meronymy or on the linguistic structures that express it. For the purposes of this thesis, I adopt the view of meronymy that Winston et al take.
2.2.4 Functionality
Unlike hyponymy and meronymy, the relation of function has not been studied in great depth. Miller sees function as a characteristic of a concept and describes it as follows: "A functional feature of a nominal concept is intended to be a description of something that instances of the concept normally do, or that is normally done with or to them" (1990, 257). Not all concepts, though, necessarily have a function or purpose. Miller asks us to consider how unnatural it is to say "the function of a canary is to fly" (1990, 257). Although flying is a normal activity for birds, it can hardly be considered its function. On the other hand, "Particularly among the human artifacts there are things that have been created for a purpose...[and can be] defined both by structure and use" (Miller: 1990, 259). This has great relevance for terminology, which deals with various fields of human endeavour which naturally involve human artifacts.
There are two "ways" that a concept can function. As mentioned above, an object can have a specific function with respect to the larger object it belongs to (such as wheels, which allow a car to move along a road). However, other objects can have a function outside of a part-whole relationship with respect to an "independent" object, but still within the same subject field. For example, an axe is used for cutting and splitting logs. Obviously the axe is not part of the log it is cutting, but the concepts axe and log are related, nonetheless, through the relation of the functionality of the axe.
Quite surprisingly, the Ahmad and Fulford (1992) study did not treat function at all. Larivière suggests that the constructions vise à, utilisé pour, sert de, and sert à express functionality in the French language (1996, 412).
I conclude that, even if function has not been greatly studied in lexical semantics, it is nonetheless important in terminology for helping understand the meaning of many concepts, particularly artifacts.
Footnote #4
Cruse views taxonomy slightly differently. He says it "may be regarded as a sub-species of hyponymy:
the taxonyms of a lexical item are a sub-set of its hyponym" (1986, 137). While the present thesis recognizes
Cruse's distinction, it will not be elaborated on here.
back to main text
Footnote #5 Ahmad and Fulford use the % and * symbols as wild cards (in a computer program) to allow for the
different inflectional variations of the probes. The % symbol stands for a single character; the * symbol stands for
any number of characters.
back to main text