<?xml version="1.0" standalone="yes"?>
<Paper uid="P86-1018">
  <Title>SEMANTICALLY SIGNIFICANT PATTERNS IN DICTIONARY DEFINITIONS *</Title>
  <Section position="4" start_page="112" end_page="113" type="metho">
    <SectionTitle>
NOUN TAXONOMY
</SectionTitle>
    <Paragraph position="0"> Noun definitions which begin with &amp;quot;Any&amp;quot; signal a taxonomic relationship between the noun being defined and a taxonomic superordinate which follows the word &amp;quot;Any.&amp;quot; One subset of the formulae beginning with &amp;quot;Any&amp;quot; has the form: &amp;quot;Any&amp;quot;-NP, where the NP can be a noun, noun phrase, or a co-ordinated noun or adjective structure.</Paragraph>
    <Paragraph position="1"> la. alkyl any univalent aliphatic, aromatic-aliphatic, or alicyclic hydrocarbon radical.</Paragraph>
    <Paragraph position="2"> b. ammunition any material used in attack or defense.</Paragraph>
    <Paragraph position="3"> c. streptococcus any coccus in chains d. nectar any delicious drink e. discord any harsh or unpleasant sound f. milkwort any herb of a genus (Pol_ygala) of the family Polygalaceae, the milkwort family In these definitions the taxonomic superordinate of the noun being defined is the head noun of the NP immediately following &amp;quot;Any&amp;quot;. The superordinate of &amp;quot;alkyl&amp;quot; is &amp;quot;radical,&amp;quot; which is the head of the co-ordinated structure following &amp;quot;Any&amp;quot; whereas the superordinate of &amp;quot;ammunition&amp;quot; is the unmodified noun &amp;quot;material.&amp;quot; Of the 97 examples of &amp;quot;Any&amp;quot;-NP only two failed to contain an overt taxonomic superordinate following &amp;quot;Any.&amp;quot; 2a. week any seven consecutive days b. couple any two persons paired together In each of these cases there is an implicit taxonomic superordinate &amp;quot;set.&amp;quot; The second frequently occurring subset of noun definitions containing &amp;quot;Any&amp;quot; begins with the following pattern: &amp;quot;Any of&amp;quot;-NP. This pattern has two principal realizations depending upon what immediately follows &amp;quot;Any of.&amp;quot; In one sub-pattern a quantifier, numeric expression, or &amp;quot;the&amp;quot; follows the initial &amp;quot;Any of&amp;quot; and begins an NP which contains the superordinate of the noun being defined. This pattern is similar to that described above for the &amp;quot;Any&amp;quot;-NP formula.  3a. doctor any of several brightly colored artificial flies b. allomorph any of two or more distinct crystalline forms of the same substance.</Paragraph>
    <Paragraph position="4"> c. elder any of various church officers The other sub-pattern expresses a biological taxonomic relationship and has the following definition structure: &amp;quot;Any of a/an&amp;quot; &lt;optional&gt; modifier taxonomic level &amp;quot;(&amp;quot;scientific name&amp;quot;)&amp;quot; &amp;quot;of&amp;quot; taxonomic superordinate either attributes or taxonomic subordinate The modifier is optional and modifies the taxonomic level of the noun being defined; the capitalized scientific name of the level follows in parenthesis; the taxonomic superordinate can be a noun or a complex NP and is the object of the second &amp;quot;of&amp;quot; in the formula; and the information following the superordinate is generally a co-ordinated structure, frequently co-ordinated NPs. Of the 901 instances of the definition-initial &amp;quot;Any of a/an&amp;quot; sequence 853, or 95 per cent, were biological definitions.</Paragraph>
    <Paragraph position="5"> 4a. ant any of a family (Formicidae) of colonial hymenopterous insects with complex social organization and various castes performing special duties.</Paragraph>
    <Paragraph position="6"> b. grass any of a large family (Gramineae) of monocotyledonous mostly herbaceous plants with jointed stems, slender sheathing leaves, and flowers borne in spikelets of bracts.</Paragraph>
    <Paragraph position="7"> c. acarld any of an order (Acarina) of arachnids including mites and ticks.</Paragraph>
    <Paragraph position="8"> d. cercis any of a small genus (Cerci s) of leguminous shrubs or low trees.</Paragraph>
    <Paragraph position="9"> e. nematode any of a class or phylum (Nematoda) of elongated cylindrical worms parasitic in animals or plants or free-living in soil or water.</Paragraph>
    <Paragraph position="10"> f. archaeornis any of a genus</Paragraph>
    <Paragraph position="12"> Jurassic toothed birds.</Paragraph>
    <Paragraph position="13"> The only sequences which break from the pattern described above are nonbiological definitions, which do not have parenthetical information following the head noun of the NP following &amp;quot;Any of a/an&amp;quot; and biological definitions where that head noun is &amp;quot;breed.&amp;quot; 5a. globulin any of a class of simple proteins (as myosin) insoluble in pure water but soluble in dilute salt solutions that occur widely in plant and animal tissues.</Paragraph>
    <Paragraph position="14"> b. rottweiler any of a breed of tall vigorous black short-haired cattle dogs.</Paragraph>
    <Paragraph position="15"> c. poland china any of an American breed of large white-marked black swine of the lard type.</Paragraph>
    <Paragraph position="16"> The definition for &amp;quot;globulin&amp;quot; illustrates that even when a nonbiological definition has a parenthesis, that parenthetical information does not immediately follow the NP following &amp;quot;Any of a/an.&amp;quot; The other definitions in (5) are instances of &amp;quot;breed&amp;quot; following &amp;quot;Any of a/an.&amp;quot; In general, when a definition begins with &amp;quot;Any of a/an&amp;quot; it is almost certainly a biological definition and that certainty is increased if the &amp;quot;Any of a/an noun&amp;quot; is immediately followed by parenthesis unless the noun of the pattern is &amp;quot;breed.&amp;quot;</Paragraph>
  </Section>
  <Section position="5" start_page="113" end_page="114" type="metho">
    <SectionTitle>
THE MEMBER-SET RELATION
</SectionTitle>
    <Paragraph position="0"> Another defining formula with an interesting resemblance to taxonomy also occurs in noun definitions. The pattern &amp;quot;A member of&amp;quot;-NP is similar to the basic organization of the &amp;quot;Any&amp;quot; definitions in that the immediate superordinate of the noun being defined is the object of the preposition &amp;quot;of&amp;quot; except in this pattern the relationship is, of course, memberset. null 6a. hand a member of a ship's crew.</Paragraph>
    <Paragraph position="1"> b. earl a member of the third grade of the British peerage ranking below a marquess and above a viscount.</Paragraph>
    <Paragraph position="2"> c. Frank a member of a West Germanic people entering the Roman provinces in A.D. 253, occupying the Netherlands and most of Gaul, and establishing themselves along the Rhine.</Paragraph>
    <Paragraph position="3"> d. republican a member of a political  party advocating republicanism e. Fox a member of an Indian people formerly living in Wisconsin.</Paragraph>
    <Paragraph position="4"> f. Episcopalian a member of an episcopal church (as the Protestant Episcopal Church).</Paragraph>
    <Paragraph position="5"> g. friar a member of a mendicant order What we have here is a generic term for any member of the speci\[led set. It Is perhaps best thought of as similar to the part-whole relation -- a hand is part of a crew, a Frank is part of a tribe, an earl is (somewhat inelegantly) part of a peerage.</Paragraph>
    <Paragraph position="6"> In our data the nouns being defined with this formula are invariably human. Of the 581 definitions which begin with &amp;quot;A member of&amp;quot; only nine define non-human nouns and two of those are anthropomorphic.</Paragraph>
    <Paragraph position="7"> 7a. Jotunn a member of a race of giants in Norse mythology b. Houyhnhnm a member of a race of horses endowed with reason in Swift's qu~li~!~ ~Y~.</Paragraph>
    <Paragraph position="8"> Why is it important to mark nouns in a lexicon as explicitly human? Many verbs can take only human subjects or objects. Also, the choice between the relative pronouns Vb9 and which depends on whether the referent is human or not.</Paragraph>
    <Paragraph position="9"> The member-set relation needs to be distinguished from another relation that classifies a specific individual as in 8a. Circe sorceress who changed Odysseus' men into swine.</Paragraph>
  </Section>
  <Section position="6" start_page="114" end_page="114" type="metho">
    <SectionTitle>
GENERIC AGENTS
</SectionTitle>
    <Paragraph position="0"> Generic agents are the typical fillers of the agent argument sot for a given verb. They are particularly valuable in understanding intersentential references or generating them. One very surprising source of definitions for human nouns is the formula &amp;quot;One that.&amp;quot; Of the 1419 examples of this pattern 694, or 49 per cent were verifiably human. That is, it was possible to determine from the definition itself or from associated definitions, such as a related verb, that the noun being defined was +human. This estimate is, therefore, conservative. It was also determined that a large portion of these definitions (30 per cent) were of occupations.</Paragraph>
    <Paragraph position="1"> 9a. goldbeater one that beats gold into gold leaf b. pollster one that conducts a poll or compiles data obtained by a poll.</Paragraph>
    <Paragraph position="2"> c. schoolmaster one that disciplines or directs.</Paragraph>
    <Paragraph position="3"> d. hatter one that makes, sells, or cleans and repairs hats.</Paragraph>
    <Paragraph position="4"> e. assassin one that murders either for hire or for fanatical motives.</Paragraph>
    <Paragraph position="5"> f. taxpayer one that pays or is liable to pay a tax g. teletyplst one that operates a teletypewriter.</Paragraph>
  </Section>
  <Section position="7" start_page="114" end_page="115" type="metho">
    <SectionTitle>
WHAT THE PARENTHESES TELL US
</SectionTitle>
    <Paragraph position="0"> The formula &amp;quot;one (..)&amp;quot; offers very different information. (This formula typically occurs somewhere in the middle of a definition, not at the beginning.) If the first word of the parenthetical information is not &amp;quot;as&amp;quot;, a definition which begins with this pattern is a biological definition. The parenthetical material is the scientific name of the noun being defined. These definitions are sub-definitions and almost invariably follow &amp;quot;esp: &amp;quot;.</Paragraph>
    <Paragraph position="1"> lOa. pimpernel any of a genus (Anagallis) of herbs of the primrose family; e~P: one (A. aryensis) whose scarlet, white, or purplish flowers close at the approach of rainy or cloudy weather.</Paragraph>
    <Paragraph position="2"> b. whelk any of numerous large marine snails (as of the genus Buccinum); esp: one (B~ undatum) much used as food in Europe.</Paragraph>
    <Paragraph position="3"> c. turnip either of two biennial herbs of the mustard family with thick roots eaten as a vegetable or fed to stock, one (Brassic@ rapa) with hairy leaves and usu. flattened roots.</Paragraph>
    <Paragraph position="4"> d. capuchin any of a genus (~ebus) of So. American monkeys; esp one (C. capuc!nas) with the hair on its crown resembling a monk's cowl.</Paragraph>
    <Paragraph position="5"> e. croton any of a genus (Crot0n) of  herbs and shrubs of the spurge famil, one (C. @lute~ia) of the Bahamas yielding cascarilla bark.</Paragraph>
    <Paragraph position="6"> f. bully tree any of several tropical American trees of the Sapodillo family; es~ one (Manilkara bid entata) that yields balata gum and heavy red timber.</Paragraph>
  </Section>
  <Section position="8" start_page="115" end_page="115" type="metho">
    <SectionTitle>
SUFFIX DEFINITIONS
</SectionTitle>
    <Paragraph position="0"> The defining pattern &amp;quot;One...(...</Paragraph>
    <Paragraph position="1"> specific /such...)&amp;quot; is an interesting sequence which is only used to define suffixes. The words &amp;quot;specific&amp;quot; and &amp;quot;such&amp;quot; signal this while at the same time indicating what semantic information should be taken from the stem to which the suffix is affixed.</Paragraph>
    <Paragraph position="2"> 11a. -er one that is a suitable object of (a specified action).</Paragraph>
    <Paragraph position="3"> b. -ate one acted upon (in a specified way).</Paragraph>
    <Paragraph position="4"> c. -morph one having (such) a form.</Paragraph>
    <Paragraph position="5"> d. -path one suffering from (such) an ailment.</Paragraph>
    <Paragraph position="6"> e. -ant one that performs (a specified action).</Paragraph>
    <Paragraph position="7"> f. -grapher one that writes about (specified) material or in a (specified) way.</Paragraph>
    <Paragraph position="8"> Examples associated with some of the definitions in (10) are &amp;quot;isomorph,&amp;quot; &amp;quot;psychopath,&amp;quot; and &amp;quot;violinist.&amp;quot; We are in the process of analyzing all instances of parenthetical &amp;quot;specified&amp;quot; and &amp;quot;such&amp;quot; to determine whether the defining formula exemplified by (10) is a general approach to the definition of affixes. Clearly, the use of parentheses is very significant, signalling an important semantic distinction.</Paragraph>
  </Section>
  <Section position="9" start_page="115" end_page="116" type="metho">
    <SectionTitle>
WHAT NOUN DEFINITIONS TELL US ABOUT VERBS
</SectionTitle>
    <Paragraph position="0"> Noun defining patterns can provide important information about specific verbs. Not surprisingly, one of these is the pattern &amp;quot;Act of Ving&amp;quot; which is an indicator of action verbs.</Paragraph>
    <Paragraph position="1"> Action verbs differ from statlve verbs in a number of important ways. Action verbs llke bite and Re rsuade can appear in imperative sentences, while statlve verbs like own and resemble cannot: Bite that man! Persuade him to go! *Own the house! *Resemble your father! Action verbs take the progressive aspect; stative verbs do not: She is biting the man.</Paragraph>
    <Paragraph position="2"> She is persuading him to go.</Paragraph>
    <Paragraph position="3"> *She is owning the house.</Paragraph>
    <Paragraph position="4"> *She is resembling your father.</Paragraph>
    <Paragraph position="5"> Action verbs can appear in a number of embedded sentences where statives cannot be used.</Paragraph>
    <Paragraph position="6"> I told her to bite the man.</Paragraph>
    <Paragraph position="7"> *I told her to own the house.</Paragraph>
    <Paragraph position="8"> In definitions the action verb appears as the gerundive object of the preposition &amp;quot;of&amp;quot; or as the present-tense verb of the subordinate clause.</Paragraph>
    <Paragraph position="9"> 12a. plumbing the act of using a plumb. b. forgiveness the act of forgiving. c. soliloquy the act of talking to oneself.</Paragraph>
    <Paragraph position="10"> d. projection the act of throwing or shooting forward.</Paragraph>
    <Paragraph position="11"> e. refund the act of refunding f. protrusion the act of protruding. g. investiture the act of ratifying or establishing in office.</Paragraph>
    <Paragraph position="12"> The examples in (11) indicate that the related verb is not always morphologically related. This pattern could, therefore, be used as a means of accessing semantically related verbs and nouns or as a tool for the construction of a semantic network.</Paragraph>
    <Paragraph position="13"> &amp;quot;The act of Ving&amp;quot; definitions have a subpattern which consists of &amp;quot;The act of Ving or the state of being &lt;adj&gt;.&amp;quot; There are not many examples of this subpattern, but in all but one instance the noun being defined, the verb and the adjective are morphologically related.</Paragraph>
    <Paragraph position="14"> 13a. adornment the act of adorning or the state of being adorned.</Paragraph>
    <Paragraph position="15"> b. popularization the act of popularizing or the state of being popularized c. nourishment the act of nourishing or the state of being nourished.</Paragraph>
    <Paragraph position="16">  d. intrusion the act of intruding or the state of being intruded.</Paragraph>
    <Paragraph position="17"> e. embodiment the act of embodying or the state of being embodied.</Paragraph>
    <Paragraph position="18"> In contrast, our data do not support the use of the corresponding formula &amp;quot;The state of being&amp;quot;-past part. for identifying stative verbs, Many instances of this pattern appear to be passives or stative use of normally non-stative verbs. This position is supported by the presence of a fair number of definitions which conjoin the two formulae.</Paragraph>
    <Paragraph position="19"> 14a. displacement the act or process of displacing: the state of being displaced.</Paragraph>
    <Paragraph position="20"> b. examination the act or process of examining: the state of being examined.</Paragraph>
    <Paragraph position="21"> c. expansion the act or process of expanding. The quality or state of being expanded.</Paragraph>
    <Paragraph position="22"> It is likely that the formula &amp;quot;The quality or state of being&amp;quot;-past part. is a stative verb indicator when it does not co-occur with &amp;quot;Act of&amp;quot; definitions. Support comes from the frequency with which that pattern alternates adjectives, which are normally stative, with the past participle.</Paragraph>
  </Section>
  <Section position="10" start_page="116" end_page="116" type="metho">
    <SectionTitle>
SELECTIONAL INFORMATION
FOR VERB DEFINITIONS
</SectionTitle>
    <Paragraph position="0"> Although the structure of verb definitions is much more limited than that of noun definitions, elements of verb definitions do provide interesting insights into collocatlonal information. One striking example of this is the use of parenthetical information which flags typical instantiations of case arguments for the verb being defined. The most consistent of these patterns is &amp;quot;To&amp;quot;-V(&lt;&amp;quot;as&amp;quot;&gt;NP) where the NP is the typical object of the verb being defined.</Paragraph>
    <Paragraph position="1"> 15a. mount to put or have (as artillery) in position.</Paragraph>
    <Paragraph position="2"> b. lay to bring forth and deposit (an egg). c. develop to subject (exposed phetographic material) to a usu. chemical treatment...</Paragraph>
    <Paragraph position="3"> We are in the process of determining how consistent the parenthetical &amp;quot;as&amp;quot; is in signalling typical case relations.</Paragraph>
  </Section>
  <Section position="11" start_page="116" end_page="116" type="metho">
    <SectionTitle>
SELECTIONAL INFORMATION FOR ADJECTIVES
</SectionTitle>
    <Paragraph position="0"> Adjective definitions differ from those of nouns and verbs in that while nouns are virtually always defined in terms of other nouns and verbs in terms of other verbs, only about 10 percent of adjectives are defined in terms of other adjectives -- the rest are related to nouns or sometimes to verbs.</Paragraph>
    <Paragraph position="1"> Furthermore, the semantic information in an adjective definition refers more to the noun (or type of noun) modified by the adjective than it does to the adjective itself. This is because an adjective, together with the noun it modifies, defines a taxonomic relationship -- or, to put it another way, denotes a feature of the thing defined in the adjective+noun phrase.</Paragraph>
    <Paragraph position="2"> For instance, we can say either that the phrase &amp;quot;big dog&amp;quot; denotes a particular kind of (the more general term) &amp;quot;dog&amp;quot;; or that it denotes a dog with the additional feature of &amp;quot;bigness&amp;quot;.</Paragraph>
    <Paragraph position="3"> A useful piece of information we would like to get from adjective definitions is selectional information -- what sort of noun the adjective can meaningfully modify. Selectional restrictions are harder to find and are largely negative -- for instance, the formula &amp;quot;containing&amp;quot; defines adjectives that do not (in the sense so defined) modify animate nouns.</Paragraph>
    <Paragraph position="4"> 10a. basic containing relatively little silica.</Paragraph>
    <Paragraph position="5"> b. normal containing neither basic hydroxyl nor acid hydrogen.</Paragraph>
    <Paragraph position="6"> The same is true of some other moderately common formulae, such as &amp;quot;consisting of&amp;quot;, &amp;quot;extending&amp;quot; and &amp;quot;causing&amp;quot;. We hope that further analysis will allow us to find more indications of selectional characteristics of adjectives.</Paragraph>
  </Section>
  <Section position="12" start_page="116" end_page="117" type="metho">
    <SectionTitle>
RECOGNIZING ACTION VS. STATIVE ADJECTIVES
</SectionTitle>
    <Paragraph position="0"> One property belonging more to adjectives themselves than to their associated nouns is an active-stative distinction similar to that found in verbs. The test for an &amp;quot;active&amp;quot; adjective is that one may use it in a statement of the form &amp;quot;they are being ---&amp;quot; or in the command &amp;quot;be .... \]&amp;quot; e.g. &amp;quot;be aggressive!&amp;quot; or &amp;quot;be good!&amp;quot;, but not *&amp;quot;be tall!&amp;quot; or *&amp;quot;be ballistic!&amp;quot; As these examples indicate, most adjectives that can be used actively can also be used  statively -- aggressiveness or goodness may be thought of as a state rather than as an action -- but not the other way around.</Paragraph>
    <Paragraph position="1"> Contrary to our expectations, the active-stative parameter of adjectives is much easier to identify in definitions than is selectlonal information. Some of the defining formulae discussed in Smith (1981) and Ahlswede (1985b) seem to be limited to statlve adjectives. &amp;quot;Of or relating to&amp;quot;, one of the most common, is one of these: fla. ballistic of or relating to ballistics or to a body in motion according to the laws of ballistics.</Paragraph>
    <Paragraph position="2"> b. literary of or relating to books.</Paragraph>
    <Paragraph position="3"> Although many adjectives defined with &amp;quot;of or relating to&amp;quot; can be used actively in other senses, they are strictly stative in the senses where this formula is used: 12a. civil of or relating to citizens &lt;~ liberties&gt;.</Paragraph>
    <Paragraph position="4"> b. peaceful of or relating to a state or time of peace.</Paragraph>
    <Paragraph position="5"> The common formula &amp;quot;being ...&amp;quot;, on the other hand, defines adjectives which at least lean toward the action end of the spectrum: 13a. natural being in accordance with or determined by nature.</Paragraph>
    <Paragraph position="6"> b. cursed being under or deserving a curse.</Paragraph>
    <Paragraph position="7"> Even such a normally stative adjective as &amp;quot;liquid&amp;quot; is relatively active in one of its senses: 14a. liquid being musical and free of harshness in sound.</Paragraph>
    <Paragraph position="8"> By no means all formulae give indications of the stative-active qualities of an adjective. A large family of formulae (&amp;quot;having&amp;quot;, &amp;quot;characterized by&amp;quot;, &amp;quot;marked by&amp;quot;, etc.) denoting attribution, are completely neutral with respect to this parameter.</Paragraph>
  </Section>
class="xml-element"></Paper>