Jumat, 02 Juli 2010

2.2. Taxonomy

Taxonomic lexical hierarchies are based on the sense relation referred to as taxonymy. Taxonymy is in fact a subtype of hyponymy since the taxonyms of a lexical item form a sub-set of its hyponyms. Taxonymy is defined as the relation of dominance in a taxonomy. Co-taxonymy then is the horizontal relation, holding between sister nodes. (Cruse 1986: 136)
A taxonomy basically requires employment of two sense relations: hyponymy between daughter-nodes and their correspondent mother-nodes, i.e. in the vertical direction, and incompatibility between sister-nodes, i.e. at the horizontal level.
A well-formed taxonomy, however, imposes additional restrictions on the semantic division of the set of lexical items. Not all hyponyms are taxonyms, and the most suitable diagnosis of taxonymy, An A is a kind/sort/type of B, sometimes fails to be appropriate to test it. Cruse (ibid., 140) suggests some approaches which make the matter clearer. There exists a strong correlation between taxonyms and so-called natural kind terms, and also between non-taxonymic hyponyms and nominal kind terms. The difference between natural kind terms and nominal kind terms consists in the nominal kind terms´ correspondence to analytic definitions which are composed of a superordinate expression and a modifier, i.e. a sort of paraphrase. A nominal kind term mentions explicitly the entailing superordinate and adds some specific feature: a female dog, the nominal kind term, paraphrases the natural kind term, a bitch. The terms liquid money or money in the form of notes and coins clearly indicate the superordinate category (i.e. money), whereas cash, the natural kind term, is specific and its meaning is inherent in the expression, but no clue is given.

An illustrative taxonomy (before the extensive taxonomic system of biological species, genera, families, etc. is presented here) is the hierarchy of language families, further divided into subfamilies (or groups), subgroups, branches, etc. This taxonomy manifests characteristic features of taxonomies, in particular:
- it consists of natural kind terms at the lowest level (names of individual languages);
- it displays the ´type/sort/kind/variety of´ relationship between a hyponym and its
hypernym;
- the members at individual levels share a certain degree of similarity, based
on genetical relatedness here (like between biological species);
- a scientific taxonomy consists of a considerably larger number of levels than
generally-known and widely-used folk taxonomies.

One more feature is also common to both the hierarchy of languages and the system of biological taxons - it is the evolutionary foundation of such taxonomic classifications. The genetical (or genealogical) classification of languages is based on the assumption that languages belonging to the same family, or subfamily, or branch, have descended from the same ancestral language, a proto-language. Therefore, apart from the enumeration of related languages (manifesting a high degree of analogous structural patterns and many similar lexical items), such a taxonomy involves an evolutionary aspect, the relation ´ancestor–descendants´. Other, non-genetical, classifications are particularly based on typology and on the geographical distribution of languages.
Two different types of a genetically-organised taxonomic hierarchy can be constructed for languages:

1. the family-tree-diagram where the ancestral language sits at the top and branches are drawn downwards and sideways to show the younger languages that have evolved from their ancestors. Such hierarchies may consist of several levels, since proto-languages which may be deemed ancestral to several descendant languages may themseleves share a common ancestor with other proto-languages spoken in the same period. The obvious drawbacks of the tree diagram are the dominance given to divergence (but languages may also converge, i.e. come closer), disregarding detail (e.g. some mid-levels in development), and failing to distinguish clearly enough between languages and dialects.

Fig. 2.1. The family-tree-diagram of the Germanic languages.

*Proto-Gmc


*Proto-WGmc *Proto-NGmc *Proto-EGmc


*Proto-AF *Proto-Neth-Gmn *Proto-WScand *Proto-EScand Gothic


English Frisian Dutch German Icelandic Norwegian Danish Swedish

Abbreviations and symbols:
* - reconstructed proto-languages
italics - extinct languages
Gmc - Germanic
WGmc - West Germanic NGmc - North Germanic EGmc - East Germanic
AF - Anglo-Frisian Neth-Gmn - Netherlandic-German WScand - West Scandinavian
EScand - East Scandinavian

Source: According to J. Lyons, Language and Linguistics. An Introduction (1981), 186.


The family-tree-diagram above (after Lyons 1981: 186) fails to include some stages and variations in the development of Germanic languages, e.g. stages in the development of English (Old English, Middle English, etc.) and their temporal parallels in other branches, as well as ancestral languages to modern Germanic languages (such as Anglo-Saxon, Old High German, Old Norse).

2. the enumeration (a list of taxons) which reflects the individual hierarchical levels (families, subfamilies, branches, groups of languages, languages) in a “report form“, running from the top to the bottom of the page and graphically distinguishing the sets of members at individual levels from each other. The highest taxon is not an ancestral language here, but a hypernym whose meaning is included in the meaning of relevant hyponyms (e.g. the hypernym Germanic languages entails English and Dutch, just as either of these languages is included in the extension of the family of Germanic languages). Another difference of an enumeration from a tree-diagram is that the evolution from ancestors to descendants is not shown; instead, the synchronic classification of related languages is used (although a group composed of extinct languages is usually added into each family).

Fig. 2.2. The enumerative (listing) classification of Germanic languages within the Indo-European family. (Slavonic languages are also presented in full extent; other groups to show the extent of the Indo-European family.)

(Language family:)
1 Indo-European family
(Branch:) (it contains groups of languages and isolated languages)
European
(Groups and isolated languages:)
1 Baltic languages
(Languages:)
Lithuanian, Latvian;
dead language: Prussian
2 Slavonic languages
West (Slavonic): Czech, Slovak, Polish, Sorbian, Kashubian;
South (Slavonic): Slovenian, Serbo-Croatian (or Serbish, Croatian),
Bulgarian, Macedonian;
East (Slavonic): Russian, Ukrainian, Belorussian;
dead languages: Old Church Slavonic, Polabian;
3 Germanic languages
North (Germanic) or Scandinavian: Icelandic, Norwegian (i.e. Bokmal and Nynorsk), Danish, Swedish, Faeroese;
West (Germanic): English, Frisian, Dutch, Flemish, Afrikaans, German, Tüütsch, Yiddish, Luxembourgish (Lëtzebuergesch);
dead languages: Gothic (Visigothic, Ostrogothic), Vandalic, Langobardian, Burgundian;
4 Celtic languages
(...)
5 Romance (or Italic) languages
(...)
6 Greek (or Hellenic)
7 Albanian
8 Armenian
9 Other extinct languages
Etruscan, Dacian, Siculian, Venedian, Frygian, etc.
Asian
10 Indic languages
(...)
11 Iranian languages
(...)
12 Dardish languages
(...)
13 Hittite-Luvian or Anatolian (all are dead)
12 Tokharian (dead)

Source: Adapted from František Čermák, Jazyk a jazykověda (2001), 64-72.

There are about 15 more families of languages (Čermák 2001: 66-71), co-hyponymous with the Indo-European family , and several isolated languages, which are not classifiable as members of any family (among them are Japanese, Korean, Basque, etc.). Sometimes geographical subgroups (such as the Ibero-Romance languages within the family of Romance languages) or geographical complexes are distinguished (such as the BalkanNS ???, including Greek, Romanian, and the regional South Slavonic languages). Some families are for their closeness sometimes also combined into one, such as the Balto-Slavonic languages, and some, based on the assumption of a common proto-language, are combined into big groups, called macroclasses, phyla or superfamilies. Thus, the hypothetical Nostratic superfamily combines the Indo-European, Hamito-Semitic, Kartvelian (or South Caucasian), Altaic (namely the Turkic group), Uralic, and Dravidian languages. Several other phyla are assumed to exist: the Afro-Asiatic, the Sino-Caucasian or Dene-Caucasian (including Basque and Sino-Tibetan languages), the Amerind, and the Austro-Asiatic or Austric (or Miao-Yao or Tai). (Čermák 2001: 72). The existence of classifications which assume hypothetical higher taxons than the generally accepted 16 or 17 families, inserting thus another level to the hierarchy, provides another evidence of a scientific origin and purpose of classification of languages, where the continuing comparison of languages and reconstruction of proto-languages may result in further grouping and re-grouping of families.


2.3. Meronomy

Meronomy is a part-whole type of branching lexical hierarchy. (Cruse 1986: 157). Such type of hierarchy is easy to find in the natural environment (parts of a human or animal body, of a plant, a glacier, etc.) or in technical disciplines (parts of an engine, device, tool, etc.). When such a hierarchy of naming units corresponding to the structure of a physical object or division of an abstract entity exists, it forms a system of meronymic relations. Meronymy is the semantic relation existing between a lexical item denoting a part and an item denoting the corresponding whole. The relation between lexical units denoting sister parts, i.e. those at the same level, is referred to as co-meronymy.
Meronomies (and taxonomic hierarchies in general) follow certain principles which determine the type of differentiation of the reality. If a whole is divided into separable, spatially or perceptually cohesive parts, these will be referred to as segmental parts. In such a division, items of a lexical hierarchy correspond to real-life objects which stand in a relation of segmental parts to the whole. An alternative approach is a division into systemic parts, which “have a greater functional unity, a greater consistency of internal constitution, but they are spatially inter-penetrating“ (ibid., 169). Divisions of this kind are not so easily perceptually accessible, but they are as valid as the former type. Every good taxonomic hierarchy must keep a constant principle of hierarchy and avoid mixing them. Thus a plant must be either divided into segmental parts, such as root, stem, leaves (further divisible into a leaf stalk or petiole, and a blade or lamina), flower, etc., or into systemic parts, such as the vascular tissue (mainly xylem and phloem), stele or vascular cylinder, cortex, stem cambium, epidermis, endodermis, photosynthetic tissue or mesophyll, and other specialised cellular systems.
Terms of both meronomic and taxonomic hierarchies denote classes of entities. However, there is a difference between the two in terms of relation between the extralinguistic reality and its reflection in lexical hierarchies. In a taxonomy, classes denoted by the terms “form a hierarchy which is more or less isomorphous with the corresponding lexical hierarchy“ (Cruse: 178). In contrast, classes denoted by the elements of a meronomy do not form a hierarchy. As Cruse (ibid., 178) states, “the hierarchical structuring of a meronomy does not originate in a hierarchy of classes.“ The relation with the reality is closer in meronomy than in taxonomy. Meronomic hierarchy is rather based on relations of individual parts to the whole.
Meronomy also is not as well-structured as taxonomy, it does not provide such well-defined and clear levels as taxonomy. There is a large number of variants and related items in meronomies, however, individual items are more clearly identifiable here than in taxonomies, which is the result of a much closer link of meronomy to the physical reality. Cruse (ibid., 178) concludes that “corresponding to a taxonomic hierarchy there is a hierarchy of classes, whereas corresponding to a part-whole hierarchy there is a class of hierarchies.“
Despite the differences, taxonomies and hierarchies are similar by that they both involve sub-division: there is a type of inclusion holding between the divided entity and the results of the division, and the relation of mutual exclusion holding between such results. Even taxonomies can be understood as a sort of part-whole relations: sub-classes can be regarded as parts of a relevant class, the whole. It is possible to say that a class referred to by its common-name label consists of its subclasses, expressed in an analogous way. This is a truly meronomic – or part-whole – relation, proving the closeness of taxonomy and meronymy.

The presence of meronomic principles alongside taxonomic ones is exemplified by most accounting hierarchies. In the terminology of accounting, taxonomies and meronomies are frequently used in combination, but they seem to have different roles: taxonomy classifies entities into categories by their function, meronomy enumerates those entities which must be added or subtracted to obtain the total sum (corresponding to an entity at a higher level) when the hierarchy is applied as a practical financial statement. The most comprehensive hierarchy, the Chart of Accounts, is formed as a nomenclature including all categories (called classes) of accounts, divided by the criteria of use into those denoting a type of property (Class 0: Intangible and tangible assets / Fixed assets, Class 1: Inventory, Class 2: Financial accounts) and those which rather denote various types of transactions and relations (Class 3: Clearing, Class 4: Capital accounts and long-term liabilities/payables, Class 5: Expenses, Class 6: Revenues, Class 7: Closing Balance Sheet accounts and off-Balance Sheet accounts, Classes 8 and 9: Managerial/Internal accounting). This highest level of classificatory division suggests a taxonomic type of hierarchy, since its items are classes, differentiated from each other and not constituting a well-defined whole (the whole would be Accounts – hence the Chart of Accounts).
The lower level of division, the one into groups of accounts, is still mostly taxonomic, but the lowest level, into individual accounts, already reveals some features of meronomy as it is often represented by an enumeration of concrete items making up a corresponding group, the whole. In real financial statements, the figures representing the value of individual parts, identical with those listed under a certain heading in the Chart of Accounts, are added up in the total figure, representing the whole on the given level of hierarchy. The problem (not linguistic, rather for compilers of such hierarchies for the purposes of various financial statements) is that not all such items – meronyms – have to be present in a certain real business entity, depending obviously on the specific type of business activity the entity is involved in. Such optionality enables to adapt the chart of accounts to the company´s relevant needs by selecting only the applicable accounts. An optional relationship between a part and a whole, usually called facultative meronymy, is frequent in selective hierarchies of both social and natural sciences and activities.


2.4. Relations similar to the part–whole relation

Meronymy is based on the existence of wholes and their constituent elements. Wholes such as groups, collections, classes etc. thus form hierarchies similar to meronymic ones. Their constituent parts are less differentiated than typical parts of meronymic hierarchies and the wholes are rather collective entities, being not so integrated as physical objects which are wholes in meronymic structures. A similar lack of differentiation applies also to elements of such collective structures. A meronym such as an eye is well-defined and limited in relation to its whole, whether it is head or body, but a member of a class, e.g. a student in relation to a university, is less definite, less singular, etc., when referred so anonymously as a member of a class. And a piece of a whole, although it is spatially limited as well and aggregation of all pieces should make up the original whole, differs from a part in the level of its autonomy. A piece must be an authentic integral component of the whole, whereas a part may be substituted for another part of the same type, such as when assembling a machine and using a specific part from several of the type that we have in stock. Other differences are that a piece, unlike a part, may have an arbitrary size and shape (the whole can be e.g. cut into large or small or irregular pieces, whereas an eye as a part is clearly defined as to its size, shape, location, etc.) and that part has a definite function it performs in the whole (cf. Cruse 1986: 158-159).
Assuming that the difference between a piece and a part in relation to their wholes has been explained, there exist the following sub-types of element–whole relations similar to meronymy:
o the group–member relation: groups are often linked with collectives of humans or animals, such as family, team, committee, jury; pack, flock, etc.). As members of groups display the property of being a part of a functional whole, the groups often have no specific lexical units to designate them.
o the class–member relation: a class is defined as “an assemblage of humans justified more by the possession of common attributes than a common purpose“ (ibid., 176). Compared with a group, a class is less cohesive as a whole and members of a class have weaker properties as parts. Unlike group nouns, class ones are usually not used in the plural form.
o the collection–member relation: in contrast with groups and classes, collections are usually inanimate. The relation in direction from member to collection is facultative, and sometimes the facultativeness functions in both directions. The members of a collection are not normally lexically distinguished (ibid., 176).
o the whole–constituent and whole–ingredient relations: these two relations manifest a significant difference: ingredients are substances which exist at the time when preparation of something starts, but which can lose their original properties or identity during the process; constituents may be created during the manufacturing or production process. The whole in such relations is usually a mass noun.
o the object–material relation: it occurs if the whole is a count noun.
o the substance–particle relation: it occurs if there is a mass-noun whole and a count-noun part (e.g. sand or salt and a grain). The lexical unit used to denote the whole can usually be used to refer to the countable particles as well as to the mass-noun wholes.

Part–whole relations allow either relationships between items on two neighbouring levels (a root is a part of a tree, a tree is a part of a forest or a park, but probably a root is not normally considered as a typical part of a forest) or between items in a chain, i.e. on different levels (a nail is a part of a finger, a finger is a part of a hand, a hand is a part of an arm, and an arm is a part of a body, but it is true at the same time that a nail is a part of a hand or an arm or a body, etc.).
Parts which are an essential feature (head as a part of (human) body) of an entity can be distinguished from those which are an optional feature of it (beard, moustache or male/female reproductive organs). Another distinction can be drawn between some items being a part (arm to a body) and those being an attribute or feature.