(305) previous ~ index ~ next

---

To: graff@unagi.cis.upenn.edu
From: Steve Lowe <steve@dragonsys.com>
Subject: Re: Question about document counts in report files
Date: Mon, 11 Jan 1999 15:15:25 -0500

---

Thanks, but that's only part of the problem. There still seems to be
a discrepency between the number of documents in tracking and in
detection. Do you have handy the total number of YES documents that
there should be in each topic (not just the change)?

Steve

Return-Path: <steve>
Date: Mon, 11 Jan 1999 12:08:31 -0500
From: Steve Lowe <steve@taz.dragonsys.com>
To: jfiscus@nist.gov, graff@unagi.cis.upenn.edu
Cc: steve@taz.dragonsys.com, jon@ishtar.dragonsys.com,
paulvm@ishtar.dragonsys.com
Subject: Question about document counts in report files


I was looking at the tracking and detection reports and am puzzled by
some apparent discrepencies in the counts. For example, in our
tracking report, for topic 88 I see 4 training + 14 correct detection
+ 5 miss = 23 on-topic stories. This agrees with what I get if I grep
for "topicid=88 level=YES" in topic_relevance.table.19990109 which
came with the results .tar archive: 23 stories. However, in our
detection report, there are only 7 reference stories for topic 88.
Also, Dave Graff's message about topic adjudication stated that 18
stories were added to the topic 88 on-topic set, so I don't see how
there can be only 7. Have I missed something?

There are other anomalies:
Topic 70, grep -> 443, trk report -> 435, det report -> 432
Topic 71, grep -> 173, trk report -> 172, det report -> 172

For reference, the above numbers are from these files in the results
tar archive:
./topic_relevance.table.19990109
./tracking/Dragon1/Dragon1_boundary_Nt=4.nwt+asr.report
./detection/Dragon1/Dragon1_boundary_DEF=10.nwt+asr.report

Steve

---

(305) previous ~ index ~ next

Last updated Wed Feb 3 10:44:21 1999