From: Daniel Ria–o (danielrr@mad.servicom.es)
Date: Sat Nov 28 1998 - 17:44:27 EST
I have been working on the implementation of a syntactical database
of Greek texts for the last six years. As I see it, the main theoretical
questions you have to resolve are the following:
a) The corpus you are going to parse.
b) The depth of the syntactical analysis
c) The model of syntactical parsing (different from:)
d) The model of syntactical description
e) The tagging scheme
f) The software implementation of the above.
Question (a) is already solved: the Greek New Testament. Now let's go for
questions b-f.
b) The depth of the syntactical analysis. I assume that in order
to do a syntactically analysed corpus, you need to have the text
morphologically parsed before, be it with a (semi-automatic) "parser" or
"manually". IMO, the best solution is adding the morphological description
to the syntactically analysed corpora, as most of modern annotated corpora
do. On the other hand, there are very interesting syntactic-semantic
questions which are normally left aside in modern annotated corpora, such
as the anaphoric relations, the syntactical categories of the words, etc,
etc. A corpus without such features is much poorer than other corpus with
them.
c) The model of syntactical parsing: You must decide first the
theoretical scheme that governs your skeleton-parsing (a dependence model,
an immediate constituents model, etc.). The chosen model limits, but not
determines, the model of syntactical description. (my choice: an immediate
constituents model).
d) The model of syntactical description: You must decide yourself
for one of the (probably hundreds of) linguistic "schools"!
e) The tagging scheme. Now we come into the theoretical-practical
problems. One thing is sure, and it is that the depth on the analysis will
exclude some tagging schemes and make some others preferable (I read a
paper about this very issue at the the 6. conference "Use of computers in
Historical and comparative linguistics" entitled "Parsing and Syntactical
Tagging of the Classical Texts. Models and Perspectives." The paper will be
published very soon if it is not already published, and I can send a
xerocopy or separata to anyone interested as soon as I have one. One
caveat: the printed version is in Spanish). It is true that most of the
schemes will allow a future user to amplify the number of categories used,
to analyse new features, etc, but the cost of re-analysing a large corpus
is enormous (and the budget for projects in humanities use to be meagre).
f) The software implementation. There are some commercial programs
for DOS, Windows and Macintosh. A good one is was implemented at the
University of Nijmegen for the "TOSCA II" corpus.. The main limitation of
commercial programs is that most of them assume that the constituents are
appear always in immediate succesion in the linguistic chain, which is
certainly not the case with ancient Greek. (When I started my project there
where only DOS programs and far from what I wanted, so that I decided to
wrote my own: in a couple of years I'll probably be releasing it for some
institutions. My project is a corpus of 80.000 words in Greek, tagged with
an annotation scheme that registers morphologic, syntactic, and semantic
features of the texts, as well as other features that belongs really to the
realm of philology, not merely linguistics. It was written in HyperCard
(!), the kind of program some people might probably think that never could
be used for such a project)
We can discuss this at length in- or off-list, if the subject
results too much off-topic. It is one of my favorite subjects!
___________________________________________________________
Daniel Ria–o Rufilanchas
c. Santa Engracia 52, 7 dcha.
28010-Madrid, Espa–a
___________________________________________________________
--- B-Greek home page: http://sunsite.unc.edu/bgreek You are currently subscribed to b-greek as: [cwconrad@artsci.wustl.edu] To unsubscribe, forward this message to leave-b-greek-329W@franklin.oit.unc.edu To subscribe, send a message to subscribe-b-greek@franklin.oit.unc.edu
This archive was generated by hypermail 2.1.4 : Sat Apr 20 2002 - 15:40:09 EDT