🚀 go-pugleaf

   -----------------------------------------------------------------
   Janet Fodor $@65<x(J $@$J$i$S$K(J Amy Weinberg $@65<x$K$h$k9V1i2q$N$*CN$i$;(J
   -----------------------------------------------------------------

            $@Aa0pEDBg3X8l3X650i8&5f=j#3#0<~G/5-G08x3+(J$@9V1i2q(J
 
               $@!V8@8l3MF@$H3X=,$K$*$1$k8@8lM}O@$NLr3d!W(J

  $@Aa0pEDBg3X8l3X650i8&5f=j#3#0<~G/$r(J$@5-G0$7$F2<5-$N$h$&$J8x3+9V1i2q$r3+(J
$@:ECW$7$^$9!#(J
  Janet Fodor $@Gn;N$O9bL>$J?4M}8@8l3X<T$G$9$,!"6aG/$OJ8(J$@=hM}$H8@8l=,F@2D(J
$@G=@-$K4X$7$F:G6a$N8@8lM}O@$H$N4XO"$G8&5f$r?J$a$F$$$^$9!#FC$K!"8=:_$O1Q(J
$@8l$HF|K\8l$NJ8(J$@=hM}$K4X$9$k<B83E*8&5f!"6g9=B$J8K!$N3X=,M}O@$J$I$K$D$$$F(J
$@$N8&5f$K=EE@$r$*$$$F$$$k$h$&$G$9!#:#2s$N9V1i(J$@$G$O!"8@8l3MF@$K4X$9$k%Q%i(J
$@%a!<%?@_Dj%b%G%k$NLdBjE@$K$D$$$F$N9M;!$r9T$J$$$^$9!#(J
  Amy Weinberg $@Gn;N$O(J$@<c<j$N7W;;8@8l3X<T!"E}8lM}O@8&5f<T$G$9$,!"(JMIT $@$N(J 
Robert Berwick  $@65<x$H$N6&Cx(J  The  Grammatical  Basis  of  Linguistic
Performance $@$G$h$/CN$i$l$F$$$^$9!#Ds>'$9$k%b%G%k$,%Q%i%a!<%?$N@_Dj$K$h$C(J
$@$F1Q8l$@$1(J$@$G$J$/F|K\8l$N=hM}$K$bBP1~$G$-$k$H$$$&$N$,:G6a$N<gD%$G$9!#:#(J
$@2s$N9V1i$G$O!"(JGB $@M}O@$K4p$E$/%Q!<%6$N8l3X(J$@650i$X$N1~MQ$K$D$$$F$*OC$$$?(J
$@$@$/M=Dj$G$9!#(J
  $@$_$J$5$^$*M6$$9g$o$;$N>e$4Mh>l2<$5$$!#(J 

                                 $@5-(J

$@<g(J    $@:E!'Aa0pEDBg3X8l3X650i8&5f=j(J
$@F|(J    $@;~!'#1#17n#1#2F|!JLZMKF|!K#1#4!'#0#0!](J$@#1#7!'#0#0(J
$@2q(J    $@>l!'Aa0pEDBg3X(J $@9q:]2q5D>l(J $@Bh;02q5D<<(J
            $@")(J169 $@?7=I6h@>Aa0pED(J 1-6-1
            ($@K\It%-%c%s%Q%9KLLg8~$+$$(J)
            $@1DCDCO2<E4El@>@~(J $@Aa0pED1X(J $@ELJb#7J,(J
            $@ETEE9S@n@~(J $@Aa0pED1X(J $@ELJb#3J,(J
            JR $@;3$N<j@~9bEDGO>l1XELJb#1#5J,(J
$@9V1iBjL\!'(J
   14:00  Prof. Amy Weinberg (University of Maryland)
            "From Linguistic Theory to Foreign Language Instruction:
             A Government-Binding Parser for Language Teaching"
           $@!V8@8lM}O@$H309q8l650i!'(JGB $@%Q!<%6(J$@$N8l3X650i$X$N1~MQ!W(J
   15:30  Prof. Janet Fodor (City University of New York)
            "Language Acquisition: Some Problems of
             Parameter Setting"
           $@!V8@8l3MF@!'%Q%i%a!<%?@_Dj%b(J$@%G%k$NLdBjE@!W(J
$@D0(J $@9V(J $@NA!'L5NA(J
$@Ld9g$;@h!'Aa0pEDBg3X(J
            $@8l3X650i8&5f=j(J $@?eLnK~(J        03-3202-4141 ext.71-5354
            $@M}9)3XIt(J $@>pJs3X2J(J $@Gr0f9nI'(J   03-3207-6528
            $@K!3XIt(J $@86ED9/Li(J              harada@Csli.stanford.edu
          $@!JEvF|$N2q>lLd9g$;$O!"Aa0pEDBg3XAm9g<uIU(J 03-3203-4141 $@$^(J$@$G!K(J

FROM LINGUISTIC THEORY TO FOREIGN LANGUAGE INSTRUCTION:
A GOVERNMENT-BINDING PARSER FOR LANGUAGE TEACHING
Amy Weinberg (University of Maryland)

   Intelligent  Computer Aided Instruction  has  held great  promise for
foreign language training but progress  has  been hampered by inflexible
designs for language processing  systems  that are the input to tutoring
systems.   This talk  presents  a  natural language processor  that  was
designed based  on principles of  the Government-  Binding  framework of
Chomsky(1981).   This system  has three  virtues  that make it the ideal
front end for  a language tutoring system:  the representation  that  it
uses is compact, allowing it to run efficiently on PC type machines that
are usually available in tutoring applications, it is based on universal
principles and so can be easily  reconfigured  to  handle  a variety  of
languages,  and  it is modular  and  so can easily ignore errors until a
student has progressed to a sufficient level of expertise to profit from
correction.   The system is also equipped  with error diagnosis feedback
which can help the student correct errors . 


LANGUAGE ACQUISITION: SOME PROBLEMS OF PARAMETER SETTING
Janet Dean Fodor (City University of New York, Graduate Center)
 
    Language  acquisition is  made difficult  by  "the  poverty  of  the
stimulus": many facts about the target language are not evidenced in the
learner's linguistic input.  One solution to this has been the PARAMETER 
MODEL.  We suppose that  the grammars  of  all  possible  languages  are 
innately  pre-established  in an infant's brain,  in the form of general 
principles,  each allowing  only  a very  small pre-determined amount of 
variation between languages.  This variation is encoded in the values of 
PARAMETERS.  A learner need only determine the right value for each 
parameter, on the basis of the input available.
   Despite its  many   advantages over  traditional   hypothesis-testing
models,  the parameter model is not without problems.  In particular, it
is vulnerable  to  ambiguity  in   the  constructions  that  trigger the
parametric "switches". For instance,  there  can  be  ambiguity  between
input that represents "core" patterns in the language, and input that is
"peripheral" or eccentric  in some way.  An idiom or  other  exceptional
construction  could look just  like the trigger  for  re-setting  a core
parameter, and so it might tempt a learner into  erroneously  re-setting
the parameter. I illustrate this danger with pronoun binding idioms such
as "I'm  going  to make me  a sandwich".  This sentence  could reset the
parameterization  of  the  binding principle  which  prohibits  a  local
antecedent for pronouns,  so learners  would overgenerate  ungrammatical
sentences such as "*I'm going to teach me Latin", "*Please control you",
etc.  In fact this does NOT happen, but our theory  of acquisition  does
not explain WHY it does not. 
   I will review several standard solutions to learnability problems and
will argue that none provides general protection against  core/periphery
ambiguities.  But  a  general solution  is  possible  if  we  revise the
learning  model so that  it admits  only  a  very  restricted  class  of
triggers.   This leads  to a very different  conception of  the relation
between core and periphery in language. 



--
  Tetsunori KOBAYASHI  (Dept.Elec.Eng., Waseda Univ.) 
      3-4-1 Okubo, Shinjuku-ku, Tokyo, 169 JAPAN
      Phone:03-3209-3211-(tone)3174 / Fax:03-3200-2567
      E-mail: koba@tk.elec.waseda.ac.jp