<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-1066">
  <Title>Parsing decomposable idioms</Title>
  <Section position="3" start_page="0" end_page="389" type="metho">
    <SectionTitle>
2 Decomposable idioms and the
</SectionTitle>
    <Paragraph position="0"> referential status of their idiom chunks hi contrast to non-compositional idioms, decomposable idioms arc; able to undergo several syntactic operations that lead to the opinion that &amp;quot;pieces of an idiom typically have identifiable meanings which combine to produce the meaning of the whole&amp;quot; (Wasow, 1982).</Paragraph>
    <Paragraph position="1"> As example, we consider the syntactic behavior of the German verbal idioms cinch Bock schieflcn (lit.: &amp;quot;shoot; a buck&amp;quot;, fig.: &amp;quot;make a mistake&amp;quot;, fig. eq.: &amp;quot;make a bloomer&amp;quot;) 2 and jmdm. einen B&amp;'cn aufbindcn (lit.: &amp;quot;tie sb. a bear on&amp;quot;, fig.: &amp;quot;tell a tall tale to sb.&amp;quot;, fig. eq.: &amp;quot;pull sb.'s leg&amp;quot;; &amp;quot;spin sb. a yarn&amp;quot;) In the following examples sew:ral modifications 2Since a high degree of language competence is necessary when judging about grammaticality of idiom constructions, we * as German native speakers choose German idioms as examples. We establish the following convention for translation: literal: literal English word-by word translation of the German idiom; figurative: English paraphrase of the figurative meaning; fig. equivalent: English idioms with an equivalent meaning.</Paragraph>
    <Paragraph position="2">  (1) Tom hat mtf der Sitzung cinch groflen \]dock geschossen. Tom has on the meeting a big |)uek shot.</Paragraph>
    <Paragraph position="3"> einige 135eke gcschossen.</Paragraph>
    <Paragraph position="4"> several bucks shot Tom made ;t big mistake on the meeting.</Paragraph>
    <Paragraph position="5"> (2) ~lbln hat in seineI,l Leben sehon Tom has in his liti~ already Tom already made several mistakes in his life.</Paragraph>
    <Paragraph position="6"> (3) :l)ic.ser~ Pcml~ hat Toni 9cschosscn.</Paragraph>
    <Paragraph position="7"> This bu(:k has '\['om shot.</Paragraph>
    <Paragraph position="8"> q.bnl made. this misl;~ke.</Paragraph>
    <Paragraph position="9"> (4) 'lTom binder; Kin\] cinch 'l*'ngla'ablich, en Tom t;ies Kin\] ;t mdmlievabh~' 'Pom |;ells Kim an mfl)eliev;tble tall tale.</Paragraph>
    <Paragraph position="10"> (5) Was fiir cinch IKiren h, at Tom Kim a'ufqcb,lmden? What for a t)em' has 'Ibm Kim ~ied..otf? Wh~t~ kind of l;a.ll tah; did Tom tell to Kim? lffiren a'@ bear on.</Paragraph>
    <Paragraph position="11">  can be found: adjectival modifications in (1, 4), (lu;mtifieation in (2), and focussing \])y (lemonstr;&gt; tive determiner (3) and by question in (5) apply C() the idioms internal NPs. It is ilnl)orta.tfl; to notice. that these oi)erations mid mo(lili(:;ttions in (I) (5) are not result of tmns or word l)lays but gratnmatieally a.nd stylistically unmarked ('onstructions. Similm&amp;quot; examples (:an be ~bttnd in other languages, too. The first German exmnple has a Dutch equivalent: een bok ~schicten, where interna\] inodifications mul quantific;~tion are 1)ossible. A french decomposable i(liom is lever u'n li&amp;n'e (lit.: &amp;quot;raise a hm:e&amp;quot;, fig.: &amp;quot;touch a (lelieate subjeet&amp;quot;); p','cndcre una cantonata (lit.:&amp;quot;take a cor-Iler') incmling &amp;quot;to make a mistake&amp;quot; is &amp;it italian one. Therefore, internal modifiability of idiolns seems not to be restri(:ted on the (;enmm lan-guage. null</Paragraph>
    <Section position="1" start_page="388" end_page="388" type="sub_section">
      <SectionTitle>
2.1 Decomposabh,. idioms are structured
entities
</SectionTitle>
      <Paragraph position="0"> It ix evident f, hat a component like bucket of a non COml)ositional idioln as kick the bucket (:annot undergo such kind of synta(:ti(: operations.</Paragraph>
      <Paragraph position="1"> Therefbre, the ineanillg of non eompositional \]dloins is seen :4.8 aAl tlnstrllctllred ('.om\]\[)lex. (,~otnt)onents like bucket which do not carry a.ny individual mea.ning are called quasi-argument.s with a non referential function (Chomsky, 1981). In opposite to this, components of decolnI)osa.ble idioms d() carry some individual meaning! &amp;quot;Of&amp;quot; (:ourse, these memfings ;~re not the literal memfings of the parts&amp;quot; is stated in (Wasow, 1982). q?hen, the questions arise, which kind of inemfing do these pm:ts carry? Which is the hidden semantic stuff of Beck or 13&amp;&amp;quot; rest)e(:tively, that is modified, inquired, quantified, and emphasized? We adopt the i)oint of view that itenis as l\]ock or /\]a~' C3dlllOt be (:onsidered as quasi argmnents but as Jigurativc. awumcnts, lhlrthermore, we follow th(~ el)in\]on that such idiomatic strings m'e 1lOt llllSLrll(;tlll'e(\[ (:()lIlplexe~,q: bill strll(:tltre(1 en-I;ities. Their sla'ueturing takes t)la(:e in t)ar;fi-M to the structuring of the literally interpwted sLring (l}urger, 1973). ()ur intuition suggests to paraI)hrase cinch I3ock schicflen t)y &amp;quot;e, inen ~%hler machen&amp;quot; (lit.: &amp;quot;make ;t misl;ake&amp;quot;) a.nd jmdm. einen 13&amp;'en aufbinden by &amp;quot;jlll(hn. eille Liigenges(:hieltte erzithlen&amp;quot;, (lit.: &amp;quot;tell a tall tale to sb.&amp;quot;), lilt ix evident for tim t)m:al)hrase and the idiom to hnve at  least the same syntactic stru(;ture as showll ill the next table.</Paragraph>
      <Paragraph position="2"> \[ -(.'\]imn \] \[ 1~()(:1~ s(:t~iet3(;i, a __H t,.,,:a: .~hoot einen ~ Fehler m~tehen a ~L ..... mistake, make i deg,'b. ~&gt;_ _l b~.a,. ,,i,,,-o~ jmdm. I eine  |I,iigengeschichte~ erziihhm o ,s~z ___\[ tall tale tell  In addition it; is importmlt that also the seman-I;i(-s of the paraphrase and the idiom can be struc-I;ured in paralM.</Paragraph>
    </Section>
    <Section position="2" start_page="388" end_page="389" type="sub_section">
      <SectionTitle>
2.2 Figuratiw,. referents of idiom chunks
</SectionTitle>
      <Paragraph position="0"> To explain this, lel; us now consider the problem Dora I;he referential point of view. We claim that individu~fl components of decolnposable illiotas can lm (:onsidered figurative arguments and that these figurative t~rgmnents lmve referents on their own.</Paragraph>
      <Paragraph position="1"> Following (Nunberg, \].978) who first discussed the referential aspect of idioms let &amp;quot;us consider that verb phn~ses 'refer' to states and ~mtivi\[;ies, and transitive verb phrases normally refer to states and activities thn,t are best identified as %pen relations' of tile form ll, xb where tt~' stands \['or \],he relation referred to 1)y the ve.rb, 'x' is a. variable for the referent of the sentence subje(:t, and 'b' stands for the referent of the object NP.&amp;quot;</Paragraph>
      <Paragraph position="3"> On this basis, an idiom is called decomposable because the situation to which it refers can be seen as an open relation Rxb.</Paragraph>
      <Paragraph position="4"> For the idiom eincn Book schicflen this means that schicflcn is a two-argument relation with a variable tbr the subject NP, the noun phrase cincn Bock referring to the concept a mistake and the verb schicflen denoting a situation where someone is acting. Extending this idea to the decomposable idiom jmdm. cinch Biiren aufbinden, it is necessary to suppose a three-argument--relation Rxyb with two open variables: x represents the sub-ject NP and y the indirect object NP. The idiom jmdm. einen B&amp;'en anfbinden is now decomposable into the noun phrase cinen Biiren, referring to a tall tale, and the verb aufbinden, referring to the activity of telling.</Paragraph>
      <Paragraph position="5"> By paraphrasing decomposable idioms, the identifiable parts of nmaning are taken into account. That means that the concept of the underlying referent, which often may be an abstract entity lacking a physical extension, should be verbalized and included into the paraphrase.</Paragraph>
      <Paragraph position="6"> Notice that in the above cases the relation between the idiom components Bock, Biir and its paraphrased referents Fchler, Liigcngcschichte is not a metaphorical one, but a conventional one.</Paragraph>
      <Paragraph position="7"> There are also decomposable idioms where decomposability is based on metaphoriocal knowledge. Besides our introspective intuition, evidence for the proposed paraphrases is found through text analyses. The strongest support comes from the everyday usage of language being observed for example in textcorpora with newspapers, literature etc. (Keil, 1.995).</Paragraph>
    </Section>
  </Section>
  <Section position="4" start_page="389" end_page="390" type="metho">
    <SectionTitle>
3 Semantic representation of
</SectionTitle>
    <Paragraph position="0"> decomposable idioms via DRT In the following, we will point out the problematic nature of meaning representation of idiomatic lmlguage with the help of DIRT (Kamp, 1993). We will show the advantages of our theoretical considerations above, that can be best illustrated by DRT already including mechanisms to handle referents. null Consider example (4) Kim bindet Tom einen unglaublichen Biiren an/(fig.:&amp;quot;Kim tells Tom an incredible tall tale&amp;quot;). DRS (6) shows the result of processing the in this case senseless -- literal reading of sentence (4) without any idiom handling procedures, a DRS (7) represents a non compositional solution: after analysing the structure syntactically, the literal meaning of the multi: word lexeme jmdm. einen Biiren aufbindcn is substituted by the &amp;quot;complex meaning&amp;quot; of the simple verb phrase as &amp;quot;jmdn. belfigen&amp;quot; (&amp;quot;lie to sb.&amp;quot;). Note that it is now a problem to represent the internal adjectival modifier incredible correctly.</Paragraph>
    <Paragraph position="1"> There is no discourse referent for that the condition incrcdible as semantic representation of the adjective unglaublich holds. Furthermore, if we want to represent the sentence Er glaubte ihr die Liigcngeschichtc (&amp;quot;He believes her the tall tale&amp;quot;) continuing example (4) --, the connection of the discourse referents cannot be made correctly as shown in DRS (8). The connection of the resumed constituent cincn unglaublichcn Biiren and the resuxning definite description dic unglaubliche Liigcngeschichtc, which definitively exists, cannot be mapped into the DRS.</Paragraph>
    <Paragraph position="2"> We claim that a more appropriate semantic representation of this idiom should respect its kind of composition and take its referents into consideration. On the base of the discussed paraphrase &amp;quot;eine Liigengeschichte erz~ihlen&amp;quot;, we offer the solution shown in (9). This representation now includes the condition incredible(z), talltalc(z), tell(z,y,z) to represent the idiom, in (10) the continuation of our sentence is shown. Reference identity between bear' and tall-tale is established by the equation u=z.</Paragraph>
    <Paragraph position="3"> What decomposable idioms concerns, now the aFor the reason of simplification, we choose English predicate names for the conditions in the DRSs, e.g. instead of logical clmlses a~s b&amp;'(x), aufbindcn(x,y,z), or bdiiqen('x,y) we present the sentence meaning with bear(x), tie-on(x,y,z), or lic-to(x,y).</Paragraph>
    <Paragraph position="4"> This way the expenditure of translation c~m be reduced in this paper.</Paragraph>
    <Paragraph position="5">  base for adequate anapllora resohltion and resolution of definite deseriplions resuming em'lier introduced discourse material is created.</Paragraph>
  </Section>
  <Section position="5" start_page="390" end_page="392" type="metho">
    <SectionTitle>
4 Used Tools: The Basic Parser
</SectionTitle>
    <Paragraph position="0"> and the Lexicon In |;lie following we introduce the tools we have used R)r parsing idiomatic sentences. We give a short description of the underlying chart parsing system (Fischer, 1995) and our idiomatic lexical database, called PIIRASEO-LEX, that we use in the sense of an additional idiom list proposed by (Wcinreich, 1969).</Paragraph>
    <Paragraph position="1"> '\['he, design of our l)arsing system was governed I)y two main goals: paralh'lism and incrcmcntality. Nevertheless different formalisms m'c used to rapresent syntactic and semantic features, having the adwmtage that for syntax as well as for semantics thc most appropriate tbrmalism can be chosen. 4 Conseqnently, to guarantee parallelism, this also requires a connection nlechanisnl t)etween these formalisms is necessary. In the following sections tim structure of I, hc parser will t)e describe(l along these lines.</Paragraph>
    <Paragraph position="2"> The grammar formalism of our systenl is an extension of the well known I)ATR-H. Synta(:tic inibrmation is encoded in feature structures. With the help of constraint equations these feature structures can be modified. The underlying unification mechanism is enriched with sequences as well as simple wflue disjunctions.</Paragraph>
    <Paragraph position="3"> For our application the se, mantic formalism is of more interest. We de, cided to adopt l)inkal's at)l)roach (13os, 1996) of I)I//F. In contrast tO Karat) l)llSs are not constructed in a to t) down fashion from a phrase structure tree, but bottom-up ltSing a version of A calculus. When coml)ining ~cah:ulus and I)RT, two different kinds of abstraction are possible. First one can abstract over a complete DRS (partial DRS) or one can abstract only over a single discourse, referent (predicative DRS). The h)llowing example shows both kind of abstraction with the A 1)RS for the indetinite tieterm|net and the noun mistake, a Feature structures are, use(t to encode the A-I)li,Ss. The main operation on A DRSs is the, fltnctional 4This is the so called co-descriptive apprvach. Using the stone fl)rmalism for syntacl;ic and scman~i(: construction is called the integrated approach, in tit(', descriptive approach they are build up sequentially.</Paragraph>
    <Paragraph position="4"> -b indicates the union of DRSs.</Paragraph>
    <Paragraph position="5"> compost|on on a partial DR S as flmctor and a predicative DRS as argument. It is implemented with the helt) of unification on the feature structures. null Our parsing engine is an active chart parser.</Paragraph>
    <Paragraph position="6"> '1)he chart edges arc marked, as usual, with cate, gory symhols. Additionally, syntactic %ature structures and )v I)RSs are attached to every edge,. 1,br the extension of active edges according to the flmdamental rule of active chart parsing, all synta('tic and semantic constraints of the resi)ective granunar rule must be satisfied.</Paragraph>
    <Paragraph position="7"> A grammar rule consists of three parts: Context free rule, s over category symbols (',onstitutc the l)aekt)onc of every grammar. They are annotate, d with equations, the solutions of which result in syntactic feature stractures. Iil the, so equations the category symbols are used as projections to mark the structures to be used. The category symbols are also used in the, senmntic operations on DllSs. For semanl;ics, besides an operator compose for functional composition, an operator id for identity is used.</Paragraph>
    <Paragraph position="8"> An interface module he, lps to connect (lifl~realt lexicons to the, t)arse, r. At the inoment, a synt;actic lexicon containing t'e, atllrC strllCtllres, a senlanl;ic lexicon with A- I)RSs and a special lexicon for idioms, calle, d PIIRASI~:o-LEx (Keil, 1994), (Kcil, 1995) are commcted to the parser.</Paragraph>
    <Paragraph position="9"> Phraseo-Lex is a computational lexicon which was specially develot)ed for idiomatic knowledge.</Paragraph>
    <Paragraph position="10"> Of all dive, rsed syntactic, semantic, and pragmatic iifformal,ion provided by 1)IIRASI.;O-LEX, we only ne, c(l for our i)url)ose h'mmata, base lexemes, (idiom t)artMt)ating lcxi(:al words: Bock, ,schicfl(,n), tim inte~rnal syntacti(: structure encoded as a syntactic tre, e, the internal sclnantic structure ellcoded as predicate argument structure and the logical tbrm.</Paragraph>
    <Paragraph position="11"> As example, we show the lexical entries of our first examplary decomt)osablc idiom.</Paragraph>
    <Paragraph position="12"> len,nta: einen Bock schieBen base lexelnes: Boek, schie\]3en internal syntactic structure:  During the parsing process this necessary idiomatic information is extracted from PHRASEO-Ll,~x and mapped into feature structures the parser can handle.</Paragraph>
    <Paragraph position="13"> 5 Processing decomposable idioms When parsing decomposable idioms with the parser described in the previous section, the following steps are taken: While initializing the ehart, it is important to control whether potential parts of an idiom are found or not. For ('.very word of a sentecne to be parsed it is checked if it is a base lexeme of an idiom. If this test was positive, an additional chart edge is inserted for every idiom the word can occur in. This edge is marked as usual, but with tile syntactic feature structure and the A DRS built from the idiomatic information of PtIRASEO-LI,',X. The feature structure of this idiom edge contains information al)out how the idiom has to be completed and its underlying syntactic structure.</Paragraph>
    <Paragraph position="14"> This information is extracted from tile PHRASEO LEX syntax tree. The following examples show the feature structures of schicflcn and Bock of ore&amp;quot; running example.</Paragraph>
    <Paragraph position="15"> case: nominativeagrm: |number:singular kperson: two stem: sehieBen_vpll3 \[head \[stem: bdegck-vple31 1 \[rest: nil Fcase: nominative\]number:singular null agrm. \]person: two gender: mast stem: boek_vpll3 vpl: \[verb:schieflen_vpll3\] The features val (for valency) respectively vpl (for verbal phraseologism) contain the information necessary to find other relevant parts for building the idiom. While in the case of verbs the feature val just contains more information than usual, namely the stems of the missing parts of the idiom, the feature vpl is used to mark idiomatic information in other syntactic feature structures. Every part of the idiom is marked with an extra ending, in our example _vpll&amp; This is due to the fact that the same words can occur in different idioms and should not be mixed up during parsing, because of the corresponding semantic structures. For example, the words Katzc and Sack occur as well in die Katzc aus dcm, Sack lassen (fig. eq. &amp;quot;let the cat out of the bag&amp;quot;) as in die Katze im Sack km@n (fig. eq. &amp;quot;buy a pig in the poke&amp;quot;). The A-DRS of the idiomatic edge already contains the literal referent of the part of the idiom they represent. This means the senmndc entry for schicficn as part of cinch Bock schicflen already contains the predicate make(z,y).</Paragraph>
    <Paragraph position="16"> PSxAy \[ make(x,y) \] PSx mistake(x) For the same reasons the A DRS for bock contains the predicate mistake(x). This information is take.n from the internal semantic structure of the idiom encoded in PIIRASEo-LEx aS shown above and translated into the A-DRS.</Paragraph>
    <Paragraph position="17"> It is important to notice that the information concerning decomposable idioms is distribute.d among all its base lexemes. Nevertheless, we only have one entry for every idiom in our idiomatic database. Only when initializing the chart, this information is spread over several edges.</Paragraph>
    <Paragraph position="18"> In the grammar, special rules must he written to handle the idiomatic edges. In these rules it must be checked whether a complete idiom can be constructed. This is done with the help of extra equations over the special features val and vpl of the idiomatic featm'e structures. The following example shows a rule connecting an object and the verb phrase of a sentence, checking if both the verb and the noun are 1)arts of the same idiom. (~  No changes were necessary neither to the chart parser itself nor to the fundamental rule. All features concerning idioms are handled in the lexicons or the grammar.</Paragraph>
    <Paragraph position="19"> The result of the parsing process are two readings of the sentence: the literal one, and the idiomatic one. The syntactic feature structures of the literal and the idiomatic reading are the same, as there is no pure syntactic difference between tile two readings. Only the semantic structures differ: one DRS represents the literal idiomatic and one the idiomatic reading.</Paragraph>
    <Paragraph position="20"> 6Feature structures and rules are reduced to a minimum in our examples to keep the structures cleat'.  This technique allows us to pm'se sentences like (1) (5) where, one part of the idiom is modified ail(1 ilot the idiom as whole. A discom'se referent for biir or hock respectively tall tale or mi,stakc is already introduced dm'ing the initialization of the chart. This referent can serve as an anchor for an I)ossible adjectival modifier as unglanblich. With the help of the rule connecting a(ljeetives and nouns (not especially written for idioms!), the predicates incredible(z) and tall-tale(z) are inserted in the DRS. This approach also works for anaphoras. The discourse referent introduce.d for B&amp;quot;ar is the. antecedent for the anapher in l;he. next sentence\]</Paragraph>
  </Section>
class="xml-element"></Paper>