<?xml version="1.0" standalone="yes"?>
<Paper uid="P85-1018">
  <Title>Using Restriction to Extend Parsing Algorithms for Complex-Feature-Based Formalisms</Title>
  <Section position="6" start_page="150" end_page="151" type="evalu">
    <SectionTitle>
5 Applications
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="150" end_page="150" type="sub_section">
      <SectionTitle>
5.1 Some Examples of the Use of the Al-
</SectionTitle>
      <Paragraph position="0"> gorithm The alg.rithnl just described liras been imph,meuted and in(',&gt;rp()rat,,&lt;l into the PATR-II Exp(,rinwntal Syst(,m at SRI Itlt,.rnali(,)lal. a gr:lmmar deveh)pment :m(l tt,~,ting envirt)nm,.))t fi,l' I'\TILII ~rammars writt(.u in Z(.t:llisl) for the Syrnl)+)li('~ 3(;(ll).</Paragraph>
      <Paragraph position="1"> The following table gives s,)me data ~ugge~t.ive of the el'feet of the restrictor on parsing etliciency, it shows the total mlnlber (,f active and passive edges added to the &lt;'hart for five sent,,ncos of up to eleven words using four different restrictors. The first allowed only category information to be ,ist,d in prodiction, thus generating th,, same l)eh:wi&lt;)r .as the un&lt;'xte:M('(} Earh,y's algorithl,,. The -,&lt;'('(,n,{ a&lt;{d,,,l su{w:tlem+,rizati+.n illf-rrllalion in a&lt;l(lili(,n t&lt;)lh(,(-:H+,~&lt;)ry: Thethird a&lt;hl-d lill.+r-gap +h,l.'ndency infornlaliou a.s well ~,&lt;+ Ihat the ~:tp pr.lif&lt;.rati&lt;,n pr-hlem wa.s r&lt;,m&lt;)ved. The lin:d restri&lt;'tor ad,lo.I v&lt;,rb form informati.n. The last c&lt;flutnn shows the p,,r('entag+, of edges that were elin,inated by using this final restrh-tor.</Paragraph>
      <Paragraph position="2">  Several facts should be kept in mind about the data above. First, for sentences with no Wh-movement or relative clauses, no gaps were ever predicted. In other words, the top-down filtering is in some sense maximal with respect to gap hypothesis. Second, the subcategorization information used in top-down filtering removed all hypotheses of constituents except for those directly subcategorized \[or. Finally, the grammar used contained constructs that would cause nontermination in the unrestricted extension of Earley's algorithm.</Paragraph>
    </Section>
    <Section position="2" start_page="150" end_page="151" type="sub_section">
      <SectionTitle>
5.2 Other Applications of Restriction
</SectionTitle>
      <Paragraph position="0"> This technique of restriction of complex-feature structures into a finite set of equivalence cla~ses can be used for a wide variety of purposes.</Paragraph>
      <Paragraph position="1"> First. parsing Mg&lt;,rithnls such ~ tile ;d~&lt;)ve (:all be modified for u~e by grain&lt;nat (ortnalintus other than P.\TR-ll. In particular, definite-clause grammars are amenable to this technique, anti it &lt;:an be IIsed to extend the Earley deduction of Pereira and Warren \[i 1 I. Pereira has use&lt;l a similar technique to improve the ellh'iency of the BI'P (bottomup h,ft-corner) parser \[71 for DCC;. I,F(; and t;PSC parsers can nlake use of the top-down filteringdevic,,a~wvll. \[:f'(; p,'tl~ot'~ n|ight be \[mill th;tl d() ll(d. r&lt;,&lt;\[11il'i. ;+. c&lt;~llt+,,,;-l'ri,~. backl&gt;&lt;.m,.</Paragraph>
      <Paragraph position="2"> * &amp;quot;;*'&lt;'(rod. rt,~ll'i&lt;'ti(.ll &lt;';tlt l)e llmt'+l If&gt; ~'llh;lllt'+' ,+l h,'r I+;~l'&gt;ill~, :dgorithuls. Ig&gt;r eX;lllll)le, tilt, ancillary fllllttic~ll to c.tlq)uto 1.1{ &lt;'l.sure-- whMi. like Ihe Earh,y alg-rithm..,itht,r du.,.+ not use feature information, or fails to terminate--,-an be modified in the same way as the Ea.rh,y I)re&lt;lict~r step to ternlinate while still using significant feature inf&lt;,rmati(m. LR parsing techniques &lt;'an therel+y I)e Ilsed f,,r ellicient par'dn~ +J conll)h,x-fe:)+ture-lmn.,&lt;l fiwnlalislun. .\l,,r(' -,l)*','ulaliv+,ly. ,'++cheme~. l'(+r s,'hed.lin~ I,I{ l&gt;:irnt.r:.-+ h~ yi..hl l,;~r.,,.-, i. l&gt;rvl &amp;quot;or+,m-e ,~r+h'r t.i:~hl I., it,,,lilie~l fi,r .',.mld.,x-f,,:lqur.-l,;r~.,,l fl)rlllaliP,.llln, alld et,'cn t1111t,&lt;\[ Iw lll+,:)+tln .d + lilt + l.(,,+.tl'ivt~+r. Finally, restriction can be ilsed ill are:~.s of i)arshlg oth+,r than top-down prediction and liltering. For inslance, in many parsing schemes, edges are indexed by a categ&lt;,ry symbol for elficient retrieval. In the case of Earley's Mgorithm.</Paragraph>
      <Paragraph position="3"> active edges can be indexed bv the category of the ,'onstituent following the dot in the dotted rule. tlowever, this again forces the primacy and atomicity of major category information. Once again, restriction can be used to solve the problem. Indexing by the restriction of the dag associated  with the need p.grmits efficient retrieval that can be tuned to the particular grammar, yet does not affect the completeness or correctness of the algorithm. The indexing can be done by discrimination nets, or specialized hashing functions akin to the partial-match retrieval techniques designed for use in Prolog implementations \[16\].</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>