Semantic and grammatical subclasses of adjectives
![3INTRODUCTION
This coursepaper is to sketch some basic principles on which a full semantic
subclasses of adjectives can be founded that will allow for constructing uniform
description templates for the individual subclasses. The underlying idea is that
there should be a one to one correspondence between the category and the template
used, like it is the case for most of the existing subclasses of nouns. For that
purpose, adjectival categories need to be defined by ‘adjectival’ expressions
(mostly with present or past participles as their head). Since this procedure
generates only a limited number of very general categories (more or less parallel to
the verb categories containing static relations), the templates are extended with
some semantic features among which the feature domain, plays an important role
since it offers the opportunity to specify on a subcategorial level the information
that most existing proposals for adjective subclasses give: the conceptual field in
which the adjective is to be situated. “On May 6, President Shavkat Mirziyoyev
chaired a meeting on measures to improve the system of teaching foreign
languages. Several areas of knowledge are determined every year in Uzbekistan,
the development of which is given priority attention. This year physics and foreign
languages have become such areas [1,34]. These conceptual fields for adjectives
appear moreover to play an important part in the construction of noun templates,
not in terms of form but also in terms of content. The ultimate aim of this proposal
is to construct a kind of template building grammar the elements of which can be
used equally for nouns, verbs and adjectives. Adjectives, a fundamental component
of language, serve to modify and describe nouns. Within the realm of adjectives,
there exist various subclasses categorized based on their semantic and grammatical
properties. These subclasses provide a framework for understanding the nuanced
ways in which adjectives contribute to meaning and sentence structure. Semantic
subclasses encompass distinctions such as qualitative, quantitative, and relational
adjectives, which convey different types of qualities, quantities, or relationships
between entities. On the other hand, grammatical subclasses involve classifications](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_3.png?v=1)
![8"I like the book that you recommended," the relative adjective 'that' specifies
which book the speaker is referring to.
Interrogative adjectives , on the other hand, are used to ask questions about a noun
and are often followed by a noun. Examples of interrogative adjectives include
'which,' 'what,' and 'whose.' Interrogative adjectives are used to ask questions about
a noun in a sentence. For example, in the sentence "Which book are you reading?"
the interrogative adjective 'which' introduces a question about the book being read.
Relative adjectives provide information about a noun while interrogative adjectives
ask questions about a noun. Both types of adjectives serve different purposes in a
sentence and help to provide more context or clarity.
There is evidence that adnominal adjectives have two syntactic sources; one as
direct modifers of the NP (or of the NP already modifed by other direct
modifers), the second as predicates (if the adjective is predicative) of a reduced
relative clause that modifes the NP plus any direct modifer adjective (cf. Sproat
and Shi’s 1990 and Cinque’s 2010 notions of direct and indirect modifcation, and
references cited there).1 Syntactically, the adjectives (more accurately Adjective
Phrases, APs) merged as direct modifers of the NP are closer to the NP than any
AP in a reduced relative clause, modify something that is still a predicate (and are
also rigidly ordered, and, arguably, a closed class):
(1)…[ DemonstrativeP … [ NumeralP … ] [ReducedRC AP ] …
[ AP … NP ]
reduced RC modifcation direct modifcationInterestingly, these two types of
syntactic modifcation appear to be associated with two distinct sets of
interpretive properties. Languages differ as to how they ultimately realize the
Merge structure in . While English does not alter it except for a relatively small
class of APs that can also occur post-nominally (to which we return), Italian
(Romance) modifes it systematically. One of the effects of these modifcations is
that the adjectives that occur pre-nominally in Italian (Romance) have only the
direct modifcation source, which provides a unique window on the type of
interpretive properties that are associated with the direct modifcation source.](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_8.png?v=1)
![9English, on the other hand, will be seen to offer a (partial) window on the
properties that are associated with the reduced relative clause source as its post-
nominal adjectives have the relative clause source as their only source. The pre-
nominal position in English and the post-nominal position in Italian are instead
structurally ambiguous between the two sources (which makes it more diffcult to
see immediately which interpretive properties are associated with each syntactic
source).The structure of the article is as follows The basic evidence of Cinque
(2010) to distinguish the two sources of adnominal APs with their associated
interpretive properties. The systematic differences between English and Italian can
be derived from a common structure of Merge (above) via different sets of
movements. The main syntactic properties of the two distinct sources of
adnominal adjectives and implications that appear to derive from the syntax of
adnominal adjectives for a number of semantic questions, such as the proper
understanding of the subsective/intersective and non-subsective usages of
adjectives and Partee’s puzzle concerning privative adjectives. Failure to
distinguish the two syntactic sources of adnominal adjectives with the associated
interpretive properties may lead one to miss the correct semantic generalizations.
The fundamental syntactic distinction between the two traditional classes of
predicative and non-predicative adjectives will also be seen to converge with the
general conclusion reached in Partee’s more recent semantic analysis of adjectives
[2,129-130].](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_9.png?v=1)
![11afraid (the child was afraid) can only occur predicatively (attributive: an afraid
child) [3,257].
A substantive adjective stands in place of a noun phrase. The noun is implied. "The
sick" is a common example of a substantive adjective, and means "sick people" (in
general). As "the sick" stands in place of a plural noun, it is treated as plural, and
takes a plural verb like "are" The sick are getting poorer. You can use substantive
adjectives to make generalisations. You can't talk about specific people or groups
of people. So "Those sick" is incorrect. Also "The sick are lying down" is incorrect
- unless you mean all sick people (in a given context). You should avoid the use of
substantive adjectives when they refer to people. Use "young people" instead of
"the young", and certainly avoid using "the blind" or "the disabled". Such
expressions reduce people to their symptoms, and are considered somewhat
impolite.
A compound adjective is a combination of two or more words which can perform
the role of an adjective in a sentence. A compound adjective is formed by adding a
noun, a present participle, or a past participle to the adjective. In some cases, the
adjective is preceded by either of these and in other cases, the adjective is followed
by either of them. Compound adjectives are mostly hyphenated. Definition of a
Compound Adjective. A compound adjective, according to the Oxford Learner’s
Dictionary, is “formed of two or more parts.” According to the Collins Dictionary,
“in grammar, a compound noun, adjective, or verb is one that is made up of two or
more words, for example fire engine, bottle-green, and force-feed.” According to
the Cambridge Dictionary, “a compound word is two or more words linked
together to produce a word with a new meaning.” So, any other part of speech that
is combined with an adjective to mean something else and describe the subject or
object in a sentence can be said to be a compound adjective.
Types of Compound Adjectives with Examples
Based on the different parts of speech or verb forms that are combined with an
adjective to form compound adjectives, we can segregate them into various types.
Implications for the semantic classification of adjectives](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_11.png?v=1)
![12Appear to derive from the two syntactic sources of adjectives and their associated
interpretive properties for the semantic classifcation of adjectives.The standard
semantic classifcation of adjectives, stemming from work by Clark, Kamp,
Parsons and Montague of the early ‘70s,14 epitomized in Partee 1995 and Kamp
and Partee 1995, considered adjectives as arranged along an ‘intersectivity’
hierarchy restricted by meaning postulates which guarantee the legitimate
inferences allowed by each class. “Adjectives are seen as forming a hierarchy
from intersective to subsective to nonsubsective, with privative adjectives an
extreme case of the nonsubsective adjectives” More recently the correctness of this
hierarchy has been called into question, at least in part. For example Partee 2003,
2007, 2009a, b, 2010 mentions Kamp’s 1975 observation that apparently
subsective adjectives like ‘tall’ or ‘big’ (tall/big for a boy =/= tall/big for a
basketball player) may be considered intersective after all, even though vague and
context dependent. The reason is that we get the same effect even if we keep the
noun constant and change other aspects of the context. My 2-year-old built a really
tall snowman. The linguistic students built a really tall snowman [4,15].
ontext dependent intersectivity has been invoked even for what Siegel
classifes as true non-intersective subsective adjectives, like skillful. Beesley also
see Kamp claims that a comparison class is always contextually determined even
for these adjectives. Imagine the context of a chess school specializing in teaching
musicians. In such a context a sentence like can mean that the violinists are good
or bad not as violinists but as chess players. We get some good violinists and some
bad violinistsThis appears to make ‘weak’ subsective adjectives like ‘tall/big’ and
‘strong’ subsective adjectives like ‘skillful/good’ closer to intersective adjectives,
with the promise of simplifying the hierarchy in. One could even try to claim
that context sensitivity is a general property of intersective adjectives. On this
view, even standard intersective adjectives like ‘blonde’, ‘bald’, ‘carnivorous’,
‘asleep’, etc., could be claimed to be context-sensitive (at least to the extent to
which the sentences in are acceptable), thus making the three cases look more
similar to each other (in their degree and type of context sensitivity](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_12.png?v=1)
![14conclusions (the non existence of a class of privative adjectives and the context-
sensitivity of adjectives of the ‘tall/big’ and of the ‘skillful/good’ type) must be
revised. Those conclusions appear to hold only for one of the two sources of
adjectives, the predicative ones (those in predicate position or merged as
predicates in an adnominal reduced relative clause). They do not hold for the
other source of adjectives, the adnominal direct modification ones (those used
non predicatively). To see this, consider the following. Recall that the pre-
nominal position of an adjective in Italian (Romance) is necessarily non-
intersective, while the post-nominal position can be either non-intersective or
intersective (depending on the adjective).18 In this light, the following interpretive
In pre-nominal position, falso ‘false, fake’ appears to have a truly privative
interpretation, one that reverses the value of the noun (not real money; not a real
friend; not a real painting), while in post-nominal position it appears to have a
subsective/intersective interpretation, indicating one of the possible forms that the
referent of the noun can take (‘money which is counterfeited’, ‘a friend who is
hypocritical’ and ‘a painting which is a copy of the original’). Interestingly, in
predicate position only the latter interpretations survive. The fact that Mandarin
Chinese only has the direct modifcation, “privative”, reading of ‘false, fake’
indirectly confirms the conclusion, drawn here on word order and interpretive
grounds, that two separate uses of ‘false’ (a direct modification and a reduced
relative clause one) should be posited. This means that privative adjectives exist
after all. And they are non-intersective, non subsective, non-predicative.
Consider now subsective adjectives of the ‘tall/big’ type, which are often taken
to always involve a comparison (whether overt or covert: ‘tall/big for a…’). As
noted above, after Higginbotham this is not always the case. In English they are
ambiguous between an absolute interpretation (‘a big thing’) and one relative to a
comparison class. As Italian (Romance) shows more clearly, the relative
interpretation is only possible with adjectives in reduced relative clauses, not
with those in direct modification. When the adjectives appear pre-nominally no
comparison class is involved [5,139].](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_14.png?v=1)
![16CHAPTER II. AN OUTLINE FOR A SEMANTIC CATEGORIZATION OF
ADJECTIVES
2.1. Lexicography and semantic categorization
The emergence of electronic monolingual dictionaries in the last two
decades has largely increased the need for a greater uniformity in lexical
descriptions, not only in their formal outline (structure, information categories), but
also on the microstructural level in the rendering of their semantic content.
Electronic organization of dictionaries in the form of databases has multiplied the
search possibilities for the user drastically. To name just a couple of things, it can
enable users to look up easily the paradigmatic environment of a word (synonyms,
near-synonyms, antonyms, hyponyms, hypernyms and other kinds of related
words) and it can help them to find words on the basis of elements of their
meaning description. These aims can of course only be realised if the semantic
descriptions of words are carried out on a systematic basis, making use of
description templates. Such templates can take different forms, according to the
purposes they serve. When used as a tool to systematise meaning descriptions in
dictionaries, it should be recommended that they have as much as possible the
same kind of structure as actual descriptions in dictionaries have. This means that
they are best organized in terms of a semantic core modified by some features,
adding further attributive information to that core. A well usable template for
dictionary descriptions would be something like
Meaning core […] Feature type 1 […] Feature type 2 […] … 2
where what is to be filled in under meaning core in the normal case is the direct
hypernym of the word defined, or a higher hypernym, or, by lack of a simple word,
a hypernymic expression and feature types are defined by the set of possible
semantic relations of the modifiers to the meaning core (such things like place,
time, manner, frequency and so on). What comes between square brackets are the
concrete realisations for a particular word. Words with the same semantic core
form together a semantic category in which at least part of the semantic behavior](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_16.png?v=1)
![17of the individual words is common. Moreover it is probable that for each category
a certain selection can be defined out of the set of possible feature types that is
applicable to its members. The kind of categorisation and the type of category
structure sketched here apply without many difficulties to noun description, as is
illustrated by a number of quite satisfactory organised on-line lexical databases of
the WordNet type, like Cornetto for Dutch and Germanet for German [6,263].
In these databases the lexicon is organised in terms of synsets, which are
paradigmatic semantic complexes, consisting of synonyms, antonyms, hyponyms
and hypernyms. For clarity’s sake, these databases do not aim at categorising
words, but by defining these synsets, they offer a workable tool for categorisation.
The hyponym-hypernym relation in fact offers a sound basis for that, the hypernym
being the category name for its set of hyponyms (internally organised in terms of
synonymy, near-synonymy, antonymy and possibly other types of relations). The
hyponyms inherit the semantic properties of the hypernym and have moreover
some supplementary semantic structure of their own. This way a quite elaborate
‘natural’ internal structure of the noun class can be built in terms of hierarchies of
classes . This is especially clear in the case of nouns denoting concrete objects. A
Lippizaner is a horse. A horse is a mammal. A mammal is an animal. Above that,
there is still the category ‘living being’, but there is no single word for that. Here
we are obliged to make use of a hypernymic expression for category definition.
Also with abstract nouns the same hyponymy-hypernymy chains do show. To take
one example from the Princeton version of WordNet: the chain excitement – joy –
emotion – feeling – state – attribute shows such an organisation of a part of the
abstracts lexicon. The reason why so much attention is paid here to a categorisation
of nouns, is not the nouns themselves, but the fact that it shows the principles of
one type of categorisation: an organisation of the lexicon in terms of itself, with
hypernyms as category names for their hyponyms and semantic properties of
words partly explicable through inheritance. This type of categorisation has the
very interesting advantage that the category name can be incorporated directly as](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_17.png?v=1)
![19the state of affairs in adjective research will be given in section. But first some
more attention must be paid to the observation just mentioned about the different
ways of categorising, by means of hyponym-hypernym relations and by means of
category definition in terms of a nominal expression. When overviewing the
literature on semantic classification of words, one can roughly discern two basic
types of classifying: the first is an organisation in terms of semantic analysis, the
second an organization in terms of the conceptual fields or themes to which the
words refer. The properties of the first type have already been discussed [7,46].
The categorization is based on the semantic decomposition of words into cores and
attributes (like in analytic definitions in dictionaries ), with the semantic core
defining the category. This type of categorization allows for some predictions
about the presence of some meaning properties by inheritance and about argument
structures. The second type of categorisation relates the word to the domain in the
world to which it belongs. In such an approach, for instance, nouns like nurse,
scalpel, bypass, patient, surgeon and so on could be taken together in a category
‘nouns related to hospitals’. This is of course a completely imaginative example,
and it is not very probable that such a category would ever be used by a
lexicographer, but it reveals the nature of this type of categorisation. The words in
such a category do not necessarily have any semantic structure in common (in the
example, there are persons, instruments and actions involved), and thus this
practice does not lead to uniformity of semantic description of the class members.
But it may have the advantage of being an easy tool for dictionary users to guide
them in some search procedures. This approach will be called henceforth the
conceptual field type of categorization. On the whole one can say that for nouns
both types of categorisation seem to coincide: the hypernym defining the class can
in most cases also be regarded as defining the conceptual sphere of the nouns
categorized. However, even here both ways of categorising do not always coincide.
This can be illustrated, among others, by food names. In WordNet a word like
steak is listed under the (indirect) hypernyms meat and food. In this case, however,](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_19.png?v=1)
![21both categorisations next to each other, treated as different information categories
in one and the same template [8,52].
A proposal
As can be guessed from section 1., the aim of this paper is to formulate a proposal
for the principles underlying a useful categorisation of adjectives in which both
basic ways of categorising are reconciled. To this aim, first two assumptions about
the basic semantic role of the adjective in phrases and in sentences have to be
discussed. The first assumption is that semantically spoken, adjectives behave
either as predicates or as predicate modifiers. They are treated as such in semantic
theories based on logic, and there are no overwhelming reasons why this practice
should not work for natural language semantics as well. If we talk about a white
flower, we mean something that is white and that is a flower, whether or not we do
that in a language containing P’s, Q’s and x’s. When we call somebody a heavy
drinker, we mean that he drinks heavily. Extensional adjectives in a natural
language can be interpreted as simple predicates, like verbs. Intensional adjectives
and privative adjectives are to be interpreted as predicate modifiers, like adverbs or
adverbial phrases. But of course adjectives do not function fully as verbs, but
rather as ‘defective’ verbs. They only share a very restricted set of semantic
valencies with the verbs. To name just one thing: adjectives have no tenses. In fact
their semantic abilities coincide more or less with those of the present and past
participles, which are also tenseless verb forms. Moreover in most of the cases
adjectives can be paraphrased by an expression having a participle as its head. So
for instance green can be paraphrased as ‘having a colour like grass’ and viral as
‘caused by a virus’. If we take these paraphrases for meaning definitions - most
dictionaries have definitions that look more or less like them – it becomes
immediately clear that some semantic patterns can be deduced from them, like
‘having the colour x’ and ‘caused by x’. These patterns have the form of and can
function as meaning categories conceived according to the semantic analysis
principle. It should also be remarked that in these paraphrases the distinction](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_21.png?v=1)
![22between a descriptive and a relational adjective becomes rather narrow. Both
paraphrases consist of a relational predicate and a second argument, the first
argument being the noun modified. The difference that remains is that in the case
of the descriptive adjective this second argument is further modified and in the
case of the real pertainism it is not. Whether this observation can be generalized
may be an interesting topic for further investigation. Languages like English have a
group of verb types which can be characterized as denoting static relations between
two or more arguments. These relations are such things as ‘be4 ‘ ‘have as a
possession’, ‘have as a property’, ‘belong to’, ‘descend from’, ‘cause’, ‘be caused
by’, ‘govern’, ‘be named’, ‘be worth’, and a few others. In the line of the semantic
analysis approach these relations constitute verbal semantic categories, even if they
sometimes only have one member.
Given this basically verbal (or at least predicative) content of the typical adjective,
and also given the widespread lexicographic practice to use participles like having,
being or caused as the head of adjective definitions, it seems plausible to assume
that basic adjective categories are directly related to some of the verbal categories
and that they share the semantic properties of those categories, however poor these
may be in most cases. In fact, often adjective categories will be ‘derived’ from
some verb category. This verbal category will in most cases consist of two-place
static predicates, while the derived adjective category will contain one-place
predicates, the category members having ‘swallowed’ one of the arguments. It may
be pretty safe to assume that by the following three basic categories having [x] as a
property, being in the situation [x] and being valued as [x]5 a pretty vast majority
of the extensional descriptive adjectives can get a first rudimentary categorisation.
They are three archetypal ways of describing things: revealing their inherent
properties, revealing external influences on them and revealing judgments on them.
For pertainisms the picture is much more diffuse: there is a whole range of
relations possible between the subject and the related nominal concept, varying
from the most general being related to to such specific relations like causing and](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_22.png?v=1)
![23descending from. A second assumption relates to the normal semantic function of
(attributive as well as predicative) adjectives: to modify nouns [9,155].
Something that is particularly elucidating on this matter is the name the adjective
bears in Dutch: bijvoeglijk naamwoord, as opposed to the Dutch name for the
noun: zelfstandig naamwoord. Let us say that the namegiving implies that they are
both considered as some kind of nouns (naamwoord), but that their status is
different. The noun is viewed as independent (zelfstandig), the adjective is not. It is
hard to translate the word bijvoeglijk since it happens to be used only in the
combination just mentioned. A translation that comes close, would be additional, a
word that suggests at the same time that what is denoted by the adjective is not the
main thing, and that it contains something above what was to be expected. So the
Dutch name for the adjective suggests that it contains nominal semantic
information that is added to the nominal information given by the noun modified.
In other words: a combination of a noun with an adjective offers semantically a
new noun with a more complex semantic structure. If butcher has as its semantic
content ‘person preparing meat for consumption’, then bald butcher can be
analysed as ‘person preparing meat for consumption and having no hair on his
head’. As we see, in this paraphrase the adjective bald has the same type of
semantic function as the inherent property ‘preparing meat for consumption’: they
both consist of a participle complemented by some arguments and modifiers, and
they both modify the semantic core of the defining formula (‘person’) and restrict
its semantic range. At this point also the semantic structure of nouns becomes
relevant for the discussion of adjectives. As is demonstrated in Wierzbicka 1985, a
lexicographical description of a noun, in order to adequately account for the use of
that word, should exceed the limits of the traditional specificans – specificatum
definition, and instead should take the form of an elaborate description of the
whole conceptual structure of the word. This stance is also adopted in the Dutch
dictionary project ANW (= Algemeen Nederlands Woordenboek) where nouns are
analysed in so-called semagrams: elaborate descriptions on the basis of fixed](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_23.png?v=1)
![24templates for each noun category. These templates consist of a semantic core (the
category name) and a number of slots for modifying expressions (modification
categories like ‘size’, ‘colour’, ‘cause’, ‘place’ and so on, which have to be filled
with concrete information for each individual noun). If, as suggested above, the
semantic behaviour of adjectives is the same as that of the ‘inherent’ semantic
modifiers of the noun, then those noun templates may prove a useful vehicle for
the conception of a categorization for adjectives. In the first place because the set
of modifier slots in the noun templates gives an overview of the ways in which
nouns can be modified, not only in terms of inherent features but also by
adjectives, as will be the hypothesis in this proposal. And in the second place, since
each nominal template disposes of a restricted set of modifier slots, adjective
categorization is linked immediately to restrictions on the nouns the category
members can be combined with. At this moment also the conceptual field strategy
for categorisation enters the scene. Notions like ‘size’, ‘colour’, ‘cause’, ‘place’ are
dimensions of semantic modification for nouns, but for the modifiers themselves
(the adjectives) they form the conceptual realms in which the individual adjectives
can be situated.
One can observe a remarkable correspondence between the set of noun modifier
slots as used in the ANW and some detailed proposals for adjective categorization,
especially those in Cornetto and GermaNet. Sometimes both systems match
exactly, sometimes several Cornetto or GermaNet categories can be taken together
in one ANW modifier slot. As a conclusion a proposal can be formulated about
semantic categorisation of adjectives, founded on two principles. First:
semantically spoken, adjectives are predicates and should be described and
categorised as such. Second: the different ways in which adjectives (or let’s be
cautious and say: the main body of adjectives) can modify nouns is a mirror image
of the modification types that can be found in the internal semantic structure of the
noun [10,123].](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_24.png?v=1)
![25Adjective categories correspond to noun modifier slots. In fact, it is imaginable
that for every adjectivenoun construction a single noun can be formed with exactly
the same semantic content. To make this a little more concrete, the proposal will be
illustrated with a few examples. Within the field of property-adjectives the feeling
category can be characterized by the definition frame Having as a property
(domain) in the domain of feeling (property) [x]. Within this frame an individual
feeling adjective like happy can be defined as: Having as a property (domain) in
the domain of feeling (property) [an experience of pleasure]. This frame can be
extended by features like cause and duration. Similar templates can be constructed
for colour, size, form, dimension, character, behavior adjectives and so on, within
the realm of property adjectives, and the same goes for situation adjectives and
evaluative adjectives. In the latter two cases the general structure is the same but,
as has been proposed above, the predicate that functions as the semantic core is
different. As an overall structure of extensional descriptive adjective templates,
something consisting of a predicate head of the type that was introduced in the
foregoing, accompanied by the necessary slots for feature types, seems to be able
to cover the large majority of the cases. Practical application will have to show
how large that majority is. The conceptual fields that were on the basis of most
proposals for adjective categorisation can be incorporated in the template as feature
type slots. Purely relational adjectives are categorised by the relational predicates
that were mentioned above. So viral and bacterial can be classified under caused
by [x], atavistic and ancestral under descending from [x]. Intensional adjectives
have to be categorized in a different way, since they do not function as modifiers to
entities but to predicates. They can be treated the same way as features in verb
templates and be categorized under the same kind of feature type slot. So for
instance heavy in heavy drinker would belong to a category to a degree [x], former
to a category in a time [x]and local to a category in a place [x]. The same may be
more or less appropriate for privative and conjectural adjectives.](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_25.png?v=1)
![27further descriptive content (for instance damn, bloody, fucking). In the category of
descriptive adjectives one can find some further distinctive oppositions, creating
subcategories [11,129].
Adjectives may be absolute or gradable (an opposition that correlates with the
presence or absence of comparative and superlative), they may have a scalar
meaning (like adjectives denoting a size) or not or they can have an extensional or
an intensional meaning (cf. Kamp 1975, where the distinguished categories are
called intersective vs. non-intersective adjectives). The latter distinction has to do
with use differences as is shown by heavy in heavy suitcase and heavy drinker,
where in the first case the thing itself is modified, and in the second case not the
person but the property ‘drinker’. Here again some parallel with the grammatical
behavior can be indicated: intensional adjectives are never used predicatively.
From Dixon (1982) on some elaborate semantic categorisations for adjectives have
been proposed. Dixon himself discerns as categories: dimension, physical property,
colour, age, value, speed, human propensity, similarity, difficulty and qualification.
A very detailed proposal is that by Hundsnurscher and Splett (1982), which has
been adopted by Cornetto as well as Germanet (Maks, Vossen, Segers and Van der
Vliet 2008). They discern 13 main classes (perceptional, spatial, temporality-
related, spatio-temporal, material-related, bodyrelated, mood-related, spirit-related,
behavior-related, social-related, quantity-related, relational, general), which are in
turn each divided into several subclasses. So for example the category of
perceptional adjectives is subdivided in lightness (bright, dark), colour (red, blue,
indigo), sound (harsh, soft), taste (sweet, sour, bitter), smell (aromatic) and surface
(soft, rough) and the category of spatial adjectives is subdivided in dimension
(long, short), direction (northern, frontal, rear), localisation (close, overseas),
origin (foreign), spatial distribution (full, sparse, overcrowded) and form (round,
rectangular). All categorisation proposals mentioned here have in common that
they appear to be of the conceptual field type.](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_27.png?v=1)
![28Categories are not defined in terms of (direct) hypernyms, but in terms of a noun or
a nominal word group denoting some domain in the world, by expressions such as
adjectives related to x, or x- related, or, as in the case of dimensional or spatial, an
adjective that is not a hypernym at all but an attribute to adjective. The problem
that has to be investigated in the next section is whether this is inevitable, and
whether an alternative type of categorisation in terms of semantic analysis and
hypernymic expressions (adjectives hardly have single words as hypernyms) is
possible [12,348].](https://docx.uz/documents/db445e2d-a1ec-4b9f-9c91-17d340591a85/page_28.png?v=1)
Semantic and grammatical subclasses of adjectives