<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-1060">
  <Title>The discourse functions of Italian subjects: a centering approach</Title>
  <Section position="4" start_page="0" end_page="352" type="metho">
    <SectionTitle>
2 Centering theory
</SectionTitle>
    <Paragraph position="0"> Centering the.ory (Grosz et al., 1986; Brennan ct al., 1987; Grosz et al., 1995) models local cohere.nee in discourse: it keet)s track of how local focus varies from one utterance to the next. (\]entering  postulates that: ~ * Each utterance Un has associated with it a set of discourse entities, the FOH.WAI{I)-LOO|&lt;IN(~ CEN-TERS or Cfs. The Cf list is ranked according to discourse salience.</Paragraph>
    <Paragraph position="1"> - The I{ACKWARD-LOOKIN(\] CENTEI{~ or Cb~ ix the t member of the Cf list that lJ,~ most centrally con- null cerns~ attd that links U~ to the previous discourse. * Finally, the PI%EFERI{.ED CENTEI{, or Cp: is the highest rauked member of the Of list. The Cp represents a prediction about die Cb of the following utterance.</Paragraph>
    <Paragraph position="2"> Transitions between two adjacent utterances U,&gt;_I and U,~ can be characterized as a function of looking backward whether Cb(Un) is the same as Cb(Un 1) - attd of looking forward whether Cb(Ur~) is the same as Cp(U,~). Table 1 illustrates the Pour transitions that are detined according to diese constraints. (Brennan et al., 1987) proposes a default ordering on transitions which correlates with discourse coherence: CONTINUE is preferred to RETAIN is prelbrred to SMOOTH-SHIFT is pre-</Paragraph>
    <Paragraph position="4"> ferred to I{()I\](\[III-SIIll.'T. 3 The saliency ordering on the Cf list, which is generally equated with grmnmatieal function, for Wes(;ern language.s is SUBJE(,T &gt; OILJI,;(',T2 &gt; ()lb .IECT 7&gt; OTI1ERS, where OTItEII.S includes pret)ositional phrases and adjuncts. (Kameyama, 1985) was the first; (,o point out that for languages such ;ts .Japanese emt)athy and topi(: lnarking alfe(:t the Cf ordering, mid t)roposed the fl)lh)wing ranking</Paragraph>
    <Paragraph position="6"> Westerll hmguages. TUFOAI ar~ll0.S l,h.'4t a IIotioll analogous to empathy arises in Wesl,ern languages as well: e.g. with I)erception verbs, it is the experichter, which is ofl;en in ot)jec.t i)osition, rather than the grammal, ieal sut)jet't, thai, shouhl be r~mked higher.</Paragraph>
    <Paragraph position="7"> Finally, centering provides an interesting fraltlt~= work for studying the functions of pronouns, as tim ot)servation that the Ct) is ofl;e.ll de.leted or pronominalized can be st&amp;ted an the following rule: Rule 1 /f some element of CI(U,, ,) is realized as a pronoun in U~, then so is Cb(U,~).</Paragraph>
    <Paragraph position="8"> This rule has been (:omt)utationally int, erpreted to individuaix: the Cb. If U,~ has: - a single pronoun, that is Cb(U,,.); (r) zero or more than one pronoml, Cb(U,~) is: C\])(U,,. 1) if (~\[)(Un 1) is realized in U,,.; otherwise the highest ranked (\]f(U,~ l) wlfieh is realized in U,, Let's apt)ly centering to the constructed eXall\]pl,, i. (2). ,n (2a) el, =- '? b,,cause the el, or a segment initial utterance is left unspecified; in (21)) the Ct) is ,lohn, as it is the only pronoun, and also the only entity t)elonging to the Cf list of (2@ a(Grosz et al., 1986; Gro,'z ct al., 1995) propose that (;tie ordo, ring on transi(;ions l)erl;~dns to sequences of transitions rather (;h3Jl to single transitions.</Paragraph>
    <Paragraph position="9"> r('alized in (21)).</Paragraph>
    <Paragraph position="10"> (2a) John is a nice guy.</Paragraph>
    <Paragraph position="12"> ii. Silo. likes him. (H,;TAIN) Cb = John, Cf = \[Mary &gt; John\] iii. She was 'with Lucy. (SMOO'FII-SIlII,'T) Cb = Mary, Cf = \[Mary &gt; Lucy\] iv. Lucy ,was with, her. (I{.OII(HI-SIlIFT) Cb = Mary, Cf = \[Lucy &gt; Mary\] \[I1 (2c).i we have a (\]ONTINUI~, as its Cb is ,\]ohn (the higlms(; entity on the Cf list of (2t))), and so is its C1). In (2c).ii, the Ct) is still Joh, n as in (2(:).i, \])ILL the (\]I) llOW is Mary~ thus we have a I,I.ETAIN. In both (2(:).iii aud (2el.iv (;Ira Cb is Mary (the only entity belonging to the Cf list; in (2t)) thai, is realized): as Mary is also (,he Cp in (2c).iii, a SMOOTII-SIlIFT occurs. Insl;('.a(1, as Lucy is I,he C I) in (2el.iv, a I{OU(HI-SIIIFT oc.ctlrs.</Paragraph>
    <Paragraph position="13"> Centering theory has ;q)pealing traits from t)oth cognitive and comi)uLational points of view. From a cognitive t)erst)ec.tive, it explains certain ph(;nOlilella of local discourse, eohe.reuce (e.g. prollolninal &amp;quot;garden l)aths'), and is supported by psycholinguisti(, e.xperiments (Gordon et al., 1993).</Paragraph>
    <Paragraph position="14"> Computationally, it is a simple mechanism, and thus it has been the basis for simple algoridm~s for anaphora resolution (Brennan el; al., 1987).</Paragraph>
    <Paragraph position="15"> Much work still remains to t)e done on (:entering. For examph;, most (tevelol)menl, so l\]u' has been based on simt)lc eonsLrucl;ed (;xamples: to aI:)ply centering to real text, issues such as how possessives and subordimtte clauses al'feet re.ferring expression resolution lnust t)e addressed.</Paragraph>
    <Paragraph position="16"> This paper is a contribution in that dire.orion.</Paragraph>
  </Section>
  <Section position="5" start_page="352" end_page="353" type="metho">
    <SectionTitle>
3 The Italian pronominal system
</SectionTitle>
    <Paragraph position="0"> Italian has two pronominal systems (CabJ)rese, 1986): weak l)ronouns, th;tt iIlust always be c.liticized to the verb (e.g. Io, le, gli - respeetiw:ly him, accusative; them, feminine accusative or her, dative.; him, dative), and strong t)ronouns (\[ui, tei,  Ioro - respee.tively tie or hiln; she or her; they or them).4 The null subject is considered part of the system of weak pronouns.</Paragraph>
    <Paragraph position="1"> Weak and strong pronouns are often in complementary distribution, as strong pronouns have to be used in prepositional phrases, e.g. per lui, for him. However, this syntactic alternation doesn't apl)ly in subject position. The choice of null versits strong pronoun de.ponds on pragmatic factors; the centering explanation offered in (Di Eugenio, 1990) goes as follows:  (3a) Typically, a nltll subject signals a CONTINUE, &amp;lid a strong pronotin a RE'FAIN or a SIIIFT.</Paragraph>
    <Paragraph position="2"> (3b) A null subject can be felicitously used in eases of RETAIN or SHIFT if in U,~ the syn null tactic context tip to and including the verbal form(s) carrying tense and / or agreement forces the null subject to refer to a particular referent and not to Cb(U,~ l)-The evidence for (3b) provided in (Di Eug(&gt; nio, 1990) derived, among others, from modal and control verb constructions, in which clitics may be clitieized to the infinitival complement of the higher verb or may climb in front of the higher verb. Wheu the clitic climbs, certain pronomiual &amp;quot;garden path&amp;quot; effects, deriving from a wrong interpretation initially assigned to the null slibject and later retracted, are avoided.</Paragraph>
  </Section>
  <Section position="6" start_page="353" end_page="355" type="metho">
    <SectionTitle>
4 Italian subjects in discourse
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="353" end_page="353" type="sub_section">
      <SectionTitle>
4.1 The corpus
</SectionTitle>
      <Paragraph position="0"> The corpus amounts to about 25 pages of text, and 12,000 words; it is composed of excerpts from two books (yon Arnim, 1989; Fallaci, 1989), a letter (Mila, 1993), a posting on the Italian bulletin board (SCI, 1994), a short story (Nichetti, 1993), and three articles from two newspapers (del Buono, 1993; Pagetti, 1993; La Nazione, 1994).</Paragraph>
      <Paragraph position="1"> The excerpts are of different lengths, with the excerI)tS from the two books being the longest.</Paragraph>
      <Paragraph position="2"> Texts were chosen to cover a variety of contemporary written Italian prose, from formal (newspaper articles about politics mid literature), to informal (posting on the Italian bulletin board), and according to the following criteria: a) ,ninimal direct speech, which has not been addressed iu cen4kui, lei, Ioro are the oblique forms of the strong system, while the nonfinative \[brms are respectively egN, ella, essi/e: in current Italian the latter forms are rarely used as the oblique forms have replaced them in subject position in my corpus there are only four occurrences of these nominative forms, and they all occur ii~ the same article (Pagetti, 1993).</Paragraph>
      <Paragraph position="3"> tering yet; b) prose that describes situations involving several animate refe, rents, bee, ause strong t)ronouns can refer only to animate referents.</Paragraph>
      <Paragraph position="4"> Table 2 shows the distrit)ution of animate third person subjects partitioned into: full NPs the numbers in parentheses refer to possessive NPs; strong pronouns; null subje(;ts I counted only those whose antecedents are not determined by contraindexing constraints (Chomsky, 1981).; other anaphors (e.g. tutte, allI~,,,) they won't be analyzed in this pat)er.</Paragraph>
    </Section>
    <Section position="2" start_page="353" end_page="354" type="sub_section">
      <SectionTitle>
4.2 Issues
</SectionTitle>
      <Paragraph position="0"> When applying (:entering to real text, one realizes that lnany issues have, llOt been solved yet. I will comment here on how deictics, possessives, and subordinate clauses affect centering.</Paragraph>
      <Paragraph position="1"> Deictics such its I, you, etc. The problem is whether they are part of the Cf list or not. I follow (Walker, 1993) in assuming that deictics are always available as part of global h)cus, and therefore are outside centering.</Paragraph>
      <Paragraph position="2"> Possessives. Table 3 includes a category marked possessive, which refers to hill NPs that inchtde a possessive adjective referring to an animate entity, such as i suoi sforzi -his cfforts. The problem is how possessives affect Cb computation and Cf ordering. While Cb computation does not appear to be affected by a possessive, that behaves like a i)ronoun, the Cf ranking ueeds to be modified. An NP of type possessive refers to two entities, the possessor Po,. and the possessed P~d. P~,d corresponds to the full NP, and thus its position in Cf is determined by the NP's grammatical flmction; as regards P,,., my working heuristics is to rank it as iinmediately preceding</Paragraph>
      <Paragraph position="4"> work, t)ut needs to be rigorously teste.d.</Paragraph>
      <Paragraph position="5"> Subordinates. Another important issue, that has not been extensively addressed ye.t - but see (Kameyama, 1997; Snri and McCoy, 1993) is how to deal with complex sentences that include coordinates and subordinates. The questions that arise concern whether there are independent Cb's and Cf lists for every clause; if not, how the Cb of the complex sentence is computed, and how semantic entities appearing in different clauses are ordered on the global Cf list.</Paragraph>
      <Paragraph position="6"> In this paper, I will loosely adopt Kmneyama's proposal (1997) that sentences containing conjtlncts and tensed adjuncts are broken down iiito a linear sequence of centering &amp;quot;units&amp;quot;, while tense-</Paragraph>
      <Paragraph position="8"/>
    </Section>
    <Section position="3" start_page="354" end_page="355" type="sub_section">
      <SectionTitle>
4.3 Centering Transitions
</SectionTitle>
      <Paragraph position="0"> Table 3 illust,r~ttes the disl,ribul;ion of referring expressions with respect I;o eenl;ering Lransil, ions.</Paragraph>
      <Paragraph position="1"> The number of full NPs in Table 3 is a.\[)(/lll; half I,heir number in ~l~bh,. 2: in 5tel;, full Nl's ofl;en inLroduce enl;iLies new 1;o Lhe discourse, in which case cenLerin g does li()l; ~ti)ply.</Paragraph>
      <Paragraph position="2"> Table 3 inehtdes two (:ohunns t;hal, don't refer t;() (:entering transitions. The ('.()tlllllu hd)eled (',ENT-ES'I' ell(lodes referring expressions that don'L refer 1,o a member of Cf(U,~ I)~ bltL l,o all eltLil, y ava.ilgd)le ill Lhc discourse. While su(:h LrmlsiLions do nol, t)elong t(i (:entering, Lhat models how (:enl, ers change h'om one cenLe.ring uniL to the nexL, Lhey cons~il, ul,e leferc.nl;ial us;rges of i~rolll)ltltS I,hat need 1;o be exI)lailmd. \[ ('.all these Lransil,ions (H,;NT-I(STAB 5 for CF, NTEIL ESTAIH,ISIIMEN'\[': })ecause such refer(,nces a.plle;tr Lo ('.sLaJ)lish the flew center of local discourse. Finally, (YI'HEI{ includes e.g. expr(;ssioils l,haL build a sel, ouL of Cb(U,~ ~) and some. other (.'nt,il;y, such a.s sia lui the sua moglic both. him and his wife.. II, is not clear how to deal with Lhese constructions within Lhe (:entering framework, and thus, I have left; lih(,m unanalyzed for the time t)eing.</Paragraph>
      <Paragraph position="3"> The resull;s are as follows. Null subje('l,s are, noL surprisingly, the m(isL frequenl, ly us('d expression 58~) for C()NTINUF,'s; 1;tie difference tmLwec.n mill Stll)je(:l,s &amp;lid all l;he other referring expressions is also st,aListically signilicant (X 2 -= 7.128, p &lt;0.01). (; Vice vers~- h CONTINIIE'8 &amp;C(:OttllL for 70~0 of null subje('ts. However, ewm full NPs can be itsed for CONTINUErs stt('h IIS;I,ges &amp;CCOIIII\[,S for 16% of (~ON'I'INUI,i'S, &amp;lid for 20~) of fldl NPs.</Paragraph>
      <Paragraph position="4">  of suggc.sLive evidence t, han as sLrong indicators, as I;hc. observal;ions in Lhe corplls~ which come Kern only 8 auLhors, are not I;otally indct)en(lent.</Paragraph>
      <Paragraph position="5"> Also, 12% of (X)N'PlNIIE',q are encoded by moans of poss('ssive NP's, and vice versa, 41% of possessive NP's are used for {X)NTINUE's.</Paragraph>
      <Paragraph position="6"> The situal;ion for ILETA1N's and SIlIl,&amp;quot;l&amp;quot;s is nol, very (:lear, as none of l,he lbur ea.t;egorics of referring expressions is t)redominml(;. All Lhese SHII,'T's are a,(:l, ually SM()()TII-SIIIFT's, i.e., l;here ;tre n() IR)U(III-SIIIFT's ~tl; all. This is n()l; surprising for null subj(~(:l,s, (;ha.L are never used for ILOII(IlI-SIlII,'T (TItr;m, 1995), however il, is puzzling for full NPs. AI)t)arently the \[l;;tli;tll wril;ers I seh'.(:Le(l a(1her(~ l,o I,t1('. (tel;roll, rankin,,; of Lr;msi(;ions, in which I~()II(HI-SIIIF'I&amp;quot;S ar{~ l;he \[easL preferred.</Paragraph>
      <Paragraph position="7"> A signilic.;mL difference in (,he usages ()f (;he \['(mr referrillg expressions regards Cli;NT-EST. \[I1. this (:a.se., full NP's are used 59% or (;he times, ;rod (;he difR;renee between full NP's, and all (;he o(;h,r Cxl)ressi(ms is signitic;ml, (X '2 -- 8.88, i ) &lt;0.01).</Paragraph>
      <Paragraph position="8"> I will now focus on the conl,rasl; t)etwet,.n zeros ;tlI(l sl,rong t)ron()ltlls, ill order to assess 1;he sLr~ttegies proposed in (3). hfiLi,tlly, (3a) zl'.yes tls(~(1 for (X)NTINUI,)~ sl;rong l)r()llt)ltltS \[()r I{I,;-TAIN &amp;lid S\[IIFq' ;q)peared nol, I,o be supportc(l, noL (wen as regards the preference for null subjecPSs for C~ONTINIIE: given Lhe numbers in Ta,ble 3, Lhe difference bel;ween zeros and strong pronouns tlsed for C()NTINUE is llOt signitieant (X ~ = 2.436, I) &lt; 0.20). This finding puzzled ,he, t)e('.~utse the usage of null sul)jeets for CONq'INI~E seems Lo be a robusl, cross-linguis~ic l)henomenon: it occurs in languages as diverse as .J;tpanese (Kameyama, 1985; Walker ('.l, al., 1994; Shim;t, 1995) and Turkish (Turan, 1995).</Paragraph>
      <Paragraph position="9"> The puzzle can lie solved by examining Lhe 1;rano siLiou preceding ~he CONTINUE ill question. '\])~&gt; ble d shows |;he dilf('.renl; possible transitions in U,, that precedes Unl I in which a (X)NTIN/IE (itcurs. The configuration in which a, CONTINIIE is preceded by st ILETAIN, whic.h \[ (:all ILET-CONT, (lilt fers from the other two be.cause, of Lhe constraint Cp(U,~) ~ Cb(Un) in l,he lUg:PAIN. This in ;t sense predicts l;h.~tI; L}te cenl;er will shift: but ill &amp; ILET- null CONT sudi prediction is not fultilled. As Table 5 shows, this has some consequences on the usage of null and strong pronouns. Compared to strong</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>