(261) previous ~ index ~ next

---

To: Jonathan Fiscus <jonathan.fiscus@nist.gov>
From: Victor Lavrenko <lavrenko@cs.umass.edu>
Subject: Re: Re-release of Evaluation Materials
Date: Fri, 11 Dec 1998 15:10:16 -0500 (EST)

---

Jon,

> In light of recent error discoveries in the index files, I'm building a
> new release of the TDT2 evaluation materials. I will be releasing them
> using two methods: FTP and on CD-ROM.

we found the following differences in the two releases of the evaluation
materils:

(1) release 1 files contain "text/" path prefix instead of "tkntext/"
(2) release 1 files contain duplicate "Non_topic_training_story" lines
(3) in release 1 files certain "Topic_training_story" docids are
repeated as "Non_topic_training_story"
(4) release 2 is missing the following "Non_topic_training_story"
stories for topic 88 (for ASR, CCAP and FDCH equally):

< # Non_topic_training_story VOA19980506.1700.0475
< # Non_topic_training_story VOA19980506.1800.0415
< # Non_topic_training_story PRI19980506.2000.0369
< # Non_topic_training_story VOA19980507.1700.0789
< # Non_topic_training_story APW19980508.0606
< # Non_topic_training_story VOA19980514.1800.0414
< # Non_topic_training_story ABC19980514.1830.0062
< # Non_topic_training_story PRI19980514.2000.0372

(1) - (3) are accounted for in the e-mail discussions of the release,
however (4) was never mentioned. Did the judgments for the listed
stories change from NO to BRIEF, or is it a glitch?
Thanks,

-- Victor
________________________________________________________
Victor Lavrenko mail: lavrenko@cs.umass.edu
(413) 545-0728 / 546-5481 http: cs.umass.edu/~lavrenko
CIIR, Computer Science dept, University of Massachusetts

---

(261) previous ~ index ~ next

Last updated Mon Dec 14 09:26:55 1998