LINGUIST List 6.160

Mon 06 Feb 1995

Qs: Japanese discourse, French "rapide", NLP & MT

Editor for this issue: <>

Directory

1. Karl Krahnke, Japanese discourse analysis
2. tatjana janicijevic, collocations of the French "rapide" (fwd)
3. Phil Bralich, More on NLP

I have a student working on a project involving text-structuring devices
(metadiscourse markers) in Japanese and English. She has found few references
to such work on Japanese. Is anyone aware of such work?

Karl Krahnke
Colorado State University
krahnkelamar.colostate.edu

Thanks in advance for any help.

I am doing a study of the use and meaning of the adjective "rapide" in
French, with particular emphasis on its collocations, for exemple: pas
rapides, mouvement rapide, geste rapide, coup d'oeil rapide. I am trying
to locate references on the semantic and distributional analysis of
'rapide'. Any suggestions about how to locate references in this area
would be greatly appreciated. I have already checked in the MLA, LLBA and
in the Bulletin analytique de la linguistique francaise. Many thanks for
any help.

Tatjana Janicijevic
tatjanaunixg.ubc.ca

To the readers:

I have recently sent out several email inquiries to various list services
that I subscribe to in an attempt to learn why we do not see more
software that can relate to the complexities of human grammar. That
is, I was trying to learn why don't we see better grammar checkers, more
translation software, and more sophisticated language tutoring software.
If there were a theory of syntax that could be programmed at all wouldn't
there also be a swarm of programs being offered by researchers and
graduate students from around the world? Wouldn't we be arguing about
the merits of "LFG-Linguist 3.1" vs "GB-Linguist 2.0." And wouldn't we
be arguing about the merits of one theories labelled bracketing versus
another's? Further, other areas that might use language interfaces such as
games and program internal "help facilities" are also quite primitive.
 Granted, the difficulties of trying to get a program to account for pro-
-blems of meaning have occupied programmers for years, but this in itself
is not sufficient enough explanation for why we do not see more results in
areas that require a program to exhibit a sophisticated understanding of
the structure of human language. Which brings me to the point of this
query.

To improve my understanding of this problem and perhaps to generate
some meaningful dialog about this problem, I would like to propose the
following list of what we should minimally expect of a program or a theory
of syntax that claims to be able to handle the structure of human language. I
would like to invite readers to add to the list or to point out which areas have
already been adequately handled. If an area is adequately handled, I would
like information on how to receive the software that is available.

I divide this list into three basic areas: 1) minimum requirements for
linguists, 2) minimum requirements for grammar checkers, foreign language
tutoring software, and other more secular uses of syntax, and 3) minimum
requirements for translation technology.

I. Minimum requirements for linguists.

A program or a theory of syntax that is to be programmed must minimally be
 able to:

 1) provide a full labelled bracketing of any string (including information
 about intermediate and maximal level nodes).
 2) provide the parts of speech of all items in a string
 3) provide the parts of a sentence (subject, direct object, etc) of a
 string
 4) provide acceptability judgements for sentences that can be parsed
 (e.g. John1 likes himself2 is parsable but unacceptable).
 5) provide judgements about Topicalized sentences, fronted wh
 questions,Tough-movement sentences, relative clauses, noun
 clauses, Cleft and Pseudo-Cleft sentences, Pied-Piping, the
 complex NP constraint, Control sentences, Parasitic Gap
 sentences, Island Conditions, the That Trace Effect, the
 coreferrence requirements of pronouns and reciprocals, and
 Strong and Weak Crossover effects.
 6) perhaps provide explanations for why particular parses fail.

II. Minimum requirements for grammar checkers, foreign language
 tutoring software, and other more secular uses of syntax.

A program or theory of syntax that is to be of value in these areas
should minimally:

 1) provide information about part of speech
 2) provide information about part of sentence (subject, object, etc)
 3) provide demonstrations of grammatical constructions e.g. make
 passivesentences from active and active from passive, make
 questions from statements, join two sentences into one using
 a variety of conjunctions, transitions and prepositions.
 4) provide judgements about errors in the production of sentences
 5) provide information about types of clauses: subordinate/main,
 adverb, noun etc.
 6) provide information about sentence type e.g. simple, complex,
 compound, etc.
 7) analyze punctuation of words, phrases, and clauses (including
 coordination).

Minimum requirements for translation technology:

 1) provide a correct parse for the subject language and provide a
 grammatically correct sentence in the target language.
 2) provide acceptability judgements and other grammar checking of
 sentences in both languages.
 3) be able to do convert some sentences in both languages: e.g.
 statement to question, active to passive and so on.

This list might be short in many ways, but I think it illustrates my point.
There is a lot more that could be happening in the realm of computational
syntax, but it is not happening. If I am overly pessimistic, I would like to
know where I can find software that can do the above. I will post all responses
to the list as a summary. Also, if anyone would like to add to this list, I
 will
keep track of the responses and post them to the list. Also, if someone can
explain what prevents these developments I would appreciate that as well.

Phil Bralich
bralichuhccux.uhcc.Hawaii.edu