noun phrase, etc.

Daniel Riaño danielrr at eresmas.net
Wed Mar 20 13:21:07 EST 2002


I have been working (for my own use) on a syntactic editor that I 
used to parse 40.000 words of Luke and Acts (as well as other authors 
of classical Greek literature). I was wandering what to do with the 
editor and the corpus after I am over with the work it was intended 
to (by the end of this year or beginning of 2003). The editor 
provides a pretty extensive marking of the text, as well as a search 
engine to recover any information you added to the text. It also 
draws trees of the parsed sentences (it can also export the results 
in the for of bracketed analysis, etc.). For every sentence, the text 
is marked as follows:

Full morphological analysis of every word (even elliptic terms!) 
including a description of the Wortbildung (prefixes, suffixes, 
infixes) of most words.
Full analysis of the immediate constituents of every sentence: for 
every constituent, **both** a tag with its syntactical function and a 
description of its constituents skeleton (in SN, SP, SV, etc terms) 
is given.
A rather extensive semantic analysis of most words (excluding high 
problematic POS like the semantics of particles; to give you a clue, 
the verbs are annotated with a media of 9 syntactical tags).
Anaphoric annotation of most elements: if you click on AU)TO/N at 
Act.Ap.1.6 you are told that its referent is )IESOU=S, that appears 
for the first time as the 13th. word of Act.Ap.1.1 (the same thing 
happens if you click on the 14th or 17th words of Act.Ap.1.1 two 
infinitives with Iesous as the subject)
The speaker of every word is [read: will be] annotated.
Several other annotations (quoted text, some idioms, etc).

For example, you can ask the program to return, activity verbs 
governing a DO (of any of several types) and a IO, only if the verb 
is placed before/after the OD, etc, etc.

Since the whole program rests on my syntactical & morphological 
analysis, etc, the output of a search will be (as Clayton wrote) not 
a function of the GNT, but a function of my analysis of it, but it 
may be better than nothing. My program is not the kind of very 
sophisticated, ultra fast and elegant software product that 
Accordance is, but it contains a lot more syntactical information (I 
like to think it has some elegance, too). It is not the product of a 
team including software developers but rather a computer by product 
of a linguist and philologist that took many years to develop.

I was wandering if there is a commercial of academic viability for 
such program and database. Of course, it will not be ready 
immediately: I have to refine much of the semantic analysis and make 
the interface more user friendly (and not just Daniel friendly), but 
if you know somebody (firma, departament) who may be interested in 
it, I'd love to hear from them. I would rather not send the program 
or the data now to third parties, but I'll be glad to perform some 
searches on request as examples of the program's utility.
-- 
~~~~~~~~~~~~~~~~~~~
Daniel Riaño Rufilanchas
Madrid, España




More information about the B-Greek mailing list