(191) previous ~ index ~ next

---

To: George Doddington <doddington@nist.gov>
From: Rich Schwartz <schwartz@bbn.com>
Subject: Cheating?
Date: Fri, 24 Sep 1999 18:42:07 -0400 (EDT)

---

It was decided to use the entire TDT-2 corpus as the test set for
this dry run, which left us no background corpus or development test on
which to tune the system. In June we were told just to cheat, since there
was no alternative and there was no other data like this. The situation
hasn't changed, and since our primary purpose for this dry run is to debug
the evaluation infrastructure, we assume that we should again use the
entire TDT-2 corpus as a background corpus and tune our results on the
test set.

Is this right?

--Rich

---

(191) previous ~ index ~ next

Last updated Tue Sep 28 10:38:17 1999