<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-2142">
  <Title>(11) Modified Modifiers Verb Adverb, Noun, Prepositional Phrase Noun Adjective, Prepositional Phra~ Adjective Adverb, Prepositional Phrase</Title>
  <Section position="2" start_page="0" end_page="0" type="metho">
    <SectionTitle>
1. The Ontological Approach
</SectionTitle>
    <Paragraph position="0"> The work on adjectives reported in tlfis paper constitutes a descriptive &amp;quot;microtheory&amp;quot; in the MikroKosmos semantic analyzer (Onyshkevych and Nirenburg 1994; and Beale et ,'d. 1995), designed to serve as a component of a lolowledge-based machine translation system (Nircnburg ct al. 1992).</Paragraph>
    <Paragraph position="1"> MikroKosmos combines findings from a variety of quasi-antonomous microtheories of language phenomena, world knowledge organization and procedural knowledge at the level of coinputer system arclfitectnrc. The basic motivation for this organization is the continued inability of the fields of lingnistics and NLP to produce a general-coverage, unified theory of trealment of language phenomena, a failure especially pronoanced in ,areas beyond computational syntax.</Paragraph>
    <Paragraph position="2"> The purpose and result of the MikroKosmos mmlysis process is the derivation of an interlingnal representation for natm'al language inlmts. The langtmge in which these representations arc expressed is called file &amp;quot;text meaning representation&amp;quot; (I'MR) langlmge, mid &amp;quot;texts&amp;quot; in dais language are called, simply, TMRs. TMR is a fr,'une-based language, where frame names typically refer to instances of ontological concepts, slot names are derived from a set of ontological properties and slot fillers ,are either elements of property value sets or pointers to concept instances.</Paragraph>
    <Paragraph position="3"> An ontology, a world model containing ilfformation about types of things, events mid properties in the world, is a necessary prerequisite for a TMR language. &amp;quot;An ontology for NLP purposes is a body of knowledge about the world (or a domain) that a) is a repository of primitive symbols used in meaning representation; b) organizes these symbols in a tangled subsumption hierarchy; * and c) further intercom~ects these symbols using a rich system of semmltic and discourse-pragmatic relations defined among the concepts&amp;quot; (Mahesh and Nirenburg 1995: 1). The function of the ontology is to supply &amp;quot;world knowledge to lexical, syntactic, and semantic processes&amp;quot; (ibid).</Paragraph>
    <Paragraph position="4"> The lexicon in MikroKosInos &amp;quot;mediates between the TMR ,and ontology&amp;quot; (Onyshkevyeh and Nirenburg 1994: 2). Lexicon entries for most open-class lexieal items represent word and plu:ase senses, which c~m be either directly mapped into ontological concepts or derived by locally (that is, in the lexicon entry itself) modifying constraints on property wducs of concepts used to specify the meaning of the given lexical item. L6xicalsenmntie information as well as clues for contextual semantic and pragmatic processing are typically located in the lexicon, adjectives being no exception. In the following section we illustrate file structnre of those parts of the lexicon entry in MikroKosmos which bear on the description of adjectival memfing.</Paragraph>
  </Section>
  <Section position="3" start_page="0" end_page="843" type="metho">
    <SectionTitle>
2. The Ontological Approach to the
</SectionTitle>
    <Paragraph position="0"> Meaning of a Typical Adjective A simple, l~rototypical case of adjectival modification is a scalar adjective, which modifies a noun both syntactically and semantically. Our microtheory associates its meaning with a region on a scale which is defined as the range of ml ontological property (cf. Carlson and Nix~enburg, 1990). Tile contribution that the adjective makes to the construction of a semantic dependency structure  (FMR) typic'ally consists of inserting its meaning (a property-value p,'fir) as a slot filler in a frame representing the me,-ming of the noun whidl this adjective syntactically modifies.</Paragraph>
    <Paragraph position="1"> Thus, in big house, big will assigu a high value as the filler of the property slot SIZE of the frame for the me,'ming of house. The range of the ontological property SIZE is a numerical and continuous scale. Each nmnerical</Paragraph>
    <Paragraph position="3"> Big, will, then, be assigned a vahte of '&gt; 0.75' value on the SIZE scale. These values ,arc a crucial pmt of the lexical mapping (LEX-MAP) from language milts to TMR units included in the senmntics (SEM-STRUC) &amp;quot;zone&amp;quot; of their lexical entries. Equally cruci,'d is the syntactic-sem,'mtic delrendency nmppiug (linking) between the syntactic-sm~eture (SYN-STRUC) and s F,M-STRUC zones, which in MikroKosmos is canied out with the help of special variables. The syntactico-scmantic information in file lexicon entry for big is as follows: t  (range (value (&gt; 0.75)) ; the value is ;in the top 25 percentile ;of the scale (relaxabte-to (value (&gt;0.6))))))))));re;laxed values are for pro;cessiug metonymies The standard tn'ocedurc for representing adiecliwd modification in TMRs is, then, to insert the scale name 1 Many zones which are actually present in the entries for these adjectives in the MikroKosmos lexi-con art., omitted from the examples.</Paragraph>
    <Paragraph position="4"> scale can be measured in an absohtte mmmer (e.g., LIN-EAR-SIZE ill feet, yards, or millimeters, or TIME ill seconds). But often natural language expressions do not refer to absohttc magnitudes but rather to abstract relative ones, as in file case of big,. We assume a 0 to 1 nulncfical range for such abstract scales. For abstract references to SIZE, lhe fillers in English e~m Ire:  and scale value for an adjective as a prolrerty-valuc pair in the frame describing the meaning of the noun the adjective meMifies. For a noun like house, whose appropilate sense (2) is directly mapped into an ontological concepl, the me,'ufing of big house will be represented as it TMR fragment shown in (3):  (3) (priwlte-home (size-attribute (value &gt; 0.75)) More complex cases of adjectival moditication are discussed in Section 4.</Paragraph>
    <Paragraph position="5"> 3. Semantic and Computational Treatment of Adjectives: Old and</Paragraph>
    <Section position="1" start_page="842" end_page="843" type="sub_section">
      <SectionTitle>
New Trends
</SectionTitle>
      <Paragraph position="0"> The literature on adjectives shows a scarcity of systematic semantic analyses or lexicx~graplfic descriptions of adjectives. Most of file linguistic scholarslfip focuses on tile taxonomies of adjectives, on file differences between the attrihutive and predicative syntactic usages as well as other syntaclic mmsformations associated with vari-.</Paragraph>
      <Paragraph position="1"> ous adjectival usages, on the qualitative/relative distinctions among adjectives, which is related to the predicative/attributive usages, and on the gradability/ comparability of qualitative adjectives (for a detailed survey, see Raskin and Nirenhurg 1995: 3~20).</Paragraph>
      <Paragraph position="2"> As colnputational semantics m~wes to large-scale syslcms serving non-toy dom~fins, the need for large leticons with entries of all lexical categories is Ireconfing increasingly acute, and the attention is turning more towards such previously neglected or avoklcd categories as the adjectives. Recently, there have appeared some first indications of lifts attetlfion.--sce, lk)r instance,  Smadja (1991), Beekwith et al. (1991), Bouillon and Viegas (1994), and Pustejovsky (1995). This research is a step in the same direction.</Paragraph>
      <Paragraph position="3"> Our analysis of adjectives, with the goal of supporting semantic analysis, shows that the issues important for adjective meaning representation are quite different from those debated in literature on adjectives. Thns, it becomes clear that the scalar~non-scalar dichotomy, and not the attributive~predicative distinction which dominates the literature, is the single most important distinction in semantic treatment of adjectives. The continuous numerical scales associated with the true scalars also render the issue of gradability and comparability rather trivial (see Raskin and Nirenburg 1995: 25-26).</Paragraph>
      <Paragraph position="4"> Another essential issue is the grain size of description. In (3) the linking attribute (SIZE) is selected rather high in the hierarchy of attributes, because in the ontology SIZE-ATTRIBUTE is the parent of such properties as LENGTH-A TT R I BUT E, WlDTtI-AqTRIBUTE, A R E A-ATTRIBUTE, WEIGHT-ATTRIBUTE, etc. If the context does not allow the analyzer to select one of those, a coarsergrain solution is preferred. In other words, we represent the meaning of big house without specifying whether big pertains to the length, width, height or area of a house. This is the result of a principled decision, based on the principle of practical effability 2(Raskin and Nirenburg 1995: 46ff), which stipulates that, in MT, the target l,'mguage should be expected to have a corresponding adjective of a comparably large grain-size.</Paragraph>
      <Paragraph position="5"> This issue has been often discussed on the example of the adjective good (cf. Katz 1972, Pustejovsky 1995).</Paragraph>
      <Paragraph position="6"> We deliberately settle on a grain size of description coarser than the most detailed semantic analysis possible (4).</Paragraph>
      <Paragraph position="7">  The finest grain-size analysis requires that a certain salient property of the modified noun is contextually selected as the one on which the meaning of the noun and that of the adjective is colmected. In our approach, the representation solution for good would be to introduce an evaluation attitude, with a high value and scoped over this property. Salient properties are, however, hard to identify formally, as is well known, for instance, in the scholarslfip on metaphor, where salience is the determining factor for the similarity dimension on whicll metaphors (,and similes) are based, It is, therefore, wise to avoid having to search for the salient property, and the principle of practical effability offers a justification for tiffs.</Paragraph>
    </Section>
  </Section>
  <Section position="4" start_page="843" end_page="845" type="metho">
    <SectionTitle>
4. Non-Property-Based Adjectival
Modification
</SectionTitle>
    <Paragraph position="0"> This section contains a brief discussion of the senmntic treatment of adjectives which cannot be reduced to the standard property-based type of adjectival modification.</Paragraph>
    <Paragraph position="1"> This discussion illustrates an important point in our approach, namely, that syntactic modification does not necessarily imply semantic modification.</Paragraph>
    <Section position="1" start_page="843" end_page="843" type="sub_section">
      <SectionTitle>
4.1 Attitudes
</SectionTitle>
      <Paragraph position="0"> Good is, of course, a scalar. Nevertheless, unlike in the case of big (2), the LEX-MAP for (4) does not contain a property-value pair that can be attached to the frame of tile modified noun like house in tile TMR. Instead, the meaning representation of good introduces an attitude on the part of the speaker with regard to the modified noun. In the TMR, the attitudes characterize the whole proposition, and thus the semantic link between the modified noun and the adjective is weakened.There are other types of adjectives which challenge the conunonsense view that the memfing of the adjective somehow &amp;quot;amalgamates&amp;quot; with the memfing of the modified noun, and most of these types are non-scalar or only marginally scalar.</Paragraph>
    </Section>
    <Section position="2" start_page="843" end_page="844" type="sub_section">
      <SectionTitle>
4.2 Temporal Adjectives
</SectionTitle>
      <Paragraph position="0"> The purely temporal knowledge in MikroKosmos is recorded with the meaning of the entire proposition, and adjective entries are not marked for it. Some temporal adjectives, of file kind that Levi presents as derived from adverbs rather than nouns (examples (1.9) in Levi 1978: 7, repeated here as 5), are analyzed in a different manner  precisely because they do not modify semantically the nouns they modify syntactically--in other words, the temporal meaning of the adjective characterizes the proposition. Thus, occasional visitor (5iii) is analyzed as a rhetorical paraphrase of visit occasionally.</Paragraph>
      <Paragraph position="1"> (5) (i) former roommate (ii) early riser (iii) occasional visitor (iv) eventual compromise</Paragraph>
    </Section>
    <Section position="3" start_page="844" end_page="844" type="sub_section">
      <SectionTitle>
4.3 Membership adjectives
</SectionTitle>
      <Paragraph position="0"> Tim membership class has been largely ignored in the literature. There has been a sporadic interest in the adjective fake (see Iw,'mska 1995--cf. Raskin 1981) because it clearly violates the silnplistic subset-forming notion of adjective meaning, such that red houses are a subset of all houses. But there are many other adjectives which use exactly the same type of lexical entry, and their similarity to each other and to fake had not been noticed before.</Paragraph>
      <Paragraph position="1"> The most typical adjectives in the member subclass are authentic (6),fake (7), and nominal (8). Many others are their synonyms and near-synonyms. The lexical entry for Otis subclass focuses on two major elements: first, wheflmr tile modified norm is a member of a certain set--anthentic and nominal members ,are but fake members are not; and, second, whether the properties of tiffs noun intersect significantly with those of tile set members--the properties of authentic members overlap with the common properties of the set meinbers on most import,'mt properties; the properties of fake members overlap with those of the set members only on unimportant properties, such as physical resembl,'mce--e.g.fake gun; and the properties of nominal members overlap more significantly with those of the set members bat not on the most important ones.</Paragraph>
      <Paragraph position="2"> The first element is represented ill a set notation: setl shows that ^$varl belongs to the set, whose typical member is denoted by a variable refseml, in tile ease of authentic and nominal but not in file case of fake. Set2 is the set of ,all properties of the members of setl; set3 is tile set of all properties of ^$var 1; set4 is, essentially, the intersection of set2 and set3.</Paragraph>
      <Paragraph position="3"> The second element is represented as the value of a saliency (importance) attitude to the intersection between the properties of the modified noun and those of the set members it is purported to belong to: the saliency v~due is 1.0 for authentic, still lfigh for nominal, and low for fake. This representation is based on the assumption that functioning as a member, which differentiates between authentic and nominal, in that the former does ,'rod tile latter does not function as a member should, is the most salient featttre, while sometlfing like physical similarity (a fake gun only looks like a gun) is the least salient one. (7) and (8) below are shown only partially, where they conWast with (6).</Paragraph>
    </Section>
    <Section position="4" start_page="844" end_page="845" type="sub_section">
      <SectionTitle>
4.4 Event-Related Adjectives
</SectionTitle>
      <Paragraph position="0"> To derive file semantic part of ,'m adjectival entry from a vel'b,'fl entry, first one must identify the case, or thematic role (such as agent, theme, beueficiary, etc.) filled by the nolal modifiexl by the adjective ill qnesfion. We illustrate this process using file lexical entries for abusive and abuse. The superentry for abuse includes at least three senses, roughly, abuse-V1 &amp;quot;insult verbally&amp;quot; abuse-V2 'violate a law or a privilege&amp;quot; and abuse-V3 'assault physically&amp;quot; ,'rod the adjective may be derived from any one of them. What is abusive is either file event (E) itself, as ill abusive speech or abusive behavior, or the agent (A) of the event, its in abusive man or abusive neighbor. AbusivelE is then tile eventive sense of the adjective formed from abuse-V1 (9), and abusive is 1A the agentive sense of the adjective in the same sense of abuse. The difference between file two is, essentially, in the position of ^$varl in the LEX-MAI' and ill the scope of atlribntion of the two attitudes inherited from file verbal entry. Natla'aUy, file adjective entries replace the verbal SYN- STRUC below wifll the standard Adj one (see  (1) above--for more data and discussion see also Raskin and Nirenburg 1996).</Paragraph>
      <Paragraph position="1"> (9) (abuse  Relalive adjectives ,are denourinal, object-related, in their meaning. The following example illustrates the connection between nominal mid adjectival meanings.</Paragraph>
      <Paragraph position="2">  As file default property connecting file modifier to the modifiexl, the MikroKosmos analyzer uses file catch-all relation PERTAIN-TO. We have identified several morn specific relations.</Paragraph>
      <Paragraph position="3"> The frst such relation is OWNED-BY, as in federal-Adjl in the sense of &amp;quot;owned by a federation.&amp;quot; Another specific relation is HAS-AS-PART, as in malignant-Adj3 in the sense of containing cancer cells. LOCATION is also a conunon relation, as in international-Adj 1, &amp;quot;taking place in a set of two or more comltries.&amp;quot; It is interesting that another sense of international utilizes the OWNED-BY property noted above, as in &amp;quot;owned by a set of two or more countries;' and yet another combines LOCATION with event-relatedness, as in &amp;quot;manufactured in a set of two or more eounlries.&amp;quot; The disambiguation mnong such multiple senses is not a simple matter, and in an unusual contraposition to the standard semantic problem of infinite polysemy, a move up, rather than down, to the undifferentiated generic meaning of an adjective like international is recommended in case of disambiguation problems. In other words, while we continue to discover more specific relations between the lexical entries of denominal adjectives mid the nouns they are derived from, file generic PERTAIN-TO property should not be discarded. This move is, again, related to file issue of grain-size of semantic description.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>