[logon] How do I score logon results from the Tanaka corpus?
Bill McNeill (UW)
billmcn at u.washington.edu
Sat Mar 28 00:54:48 CET 2009
I have been able to run a logon experiment on the Tanaka Japanese corpus and
generate a score file in the profile, however, I am unable to evaluate the
accuracy. I'm not sure what I'm doing wrong.
I have run a single parse ranking experiment on profile ja_6 in the Tanaka
corpus, which on my machine is under
/home/billmcn/corpora/Tanaka/tc-070707
I now have a
[tanaka_ja_6] GP[0] +PT -LEX CW[] -AE NS[0] NT[] -NB LM[0] FT[:::1] RS[]
MM[tao_lmvm] MI[5000] RT[1.0e-6] AT[1.0e-20] VA[1.0e+4] PC[100]
in this directory. This directory contains non-zero relations, item, score,
and fold files. The contents of the score file look correct.
I have also a full logon tree checked out under:
/home/billmcn/logon/lingo
I want to run summarize-folds on the output of my experiment and get
accuracy numbers. If I feed the following .lisp script into load
(setf *tsdb-home* "/home/billmcn/logon/lingo/redwoods/tsdb/home")
(summarize-folds :output "/home/billmcn/temp/tanaka_ja_6.results" :pattern
"\\[tanaka_ja_6\\]")
it runs for a couple minutes and generates an empty tanaka_ja_6.results
file. The STDOUT from load all looks correct to me. (This is attached.)
Any suggestions on what I'm doing wrong?
--
Bill McNeill
http://staff.washington.edu/billmcn/index.shtml
Sent from: Seattle Washington United States.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.emmtee.net/archives/logon/attachments/20090327/72acc159/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tanaka-score.stdout.gz
Type: application/x-gzip
Size: 3042 bytes
Desc: not available
URL: <http://lists.emmtee.net/archives/logon/attachments/20090327/72acc159/attachment.gz>
More information about the logon
mailing list