[b-greek] Re: noun phrase, etc.

From: c stirling bartholomew (cc.constantine@worldnet.att.net)
Date: Thu Mar 14 2002 - 22:26:21 EST

on 3/14/02 5:47 PM, c stirling bartholomew wrote:

> If not then what you will need to do if you want to be able to locate every
> "Noun Phrase" in your electronic GNT is have Dale W. go and tag each "Noun
> Phrase" manually. What you will end up with is a list of what Dale decided
> were "Noun Phrases." Simple, no? It is going to be a function of your GNT
> database, not a function of your search engine.

The off list mail indicates that I should issue a clarification.

The request for a "better search engine" is one that I have made myself
several times on this list. I appreciate the request. It is a valid and
worthwhile desire to want to be able to search on things other than
morphology in the GNT, however there are problems in implementation. Big

Not very many years ago I was yelling and screaming at Gramcord to come up
with some semantic functional tags for their GNT so we could search on
things more interesting than just morphology. However, after Paul Miller
"clarified" the issues for me and I did a little more thinking about the
problem I realized that my wish list for "high level" search criteria was
more than just a little bit utopian.

One problem is that "high level" search criteria are very closely linked to
a particular language model, and these models go in and out of style at a
rapid rate. They become obsolete in five to ten years. By the time a model
is chosen, the specification for the database tags is worked out and the
data base is tagged, the model itself will be a curio of linguist history.

The best way to get some appreciation for the magnitude of the problem is to
do some rule writing using a good morph-tagged GNT and see how far you get.
This exercise will instill vast quantities of humility when you discover
that even a simple thing like a noun phrase is very difficult to nail down.


Clayton Stirling Bartholomew
Three Tree Point
P.O. Box 255 Seahurst WA 98062

B-Greek home page: http://metalab.unc.edu/bgreek
You are currently subscribed to b-greek as: [jwrobie@mindspring.com]
To unsubscribe, forward this message to leave-b-greek-327Q@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 - 15:37:21 EDT