Nocem and Spam

Hi,
the big spam waves are over and that's a good thing.
Consulting the excellent <https://www.novabbs.com/SEARCH/search_nocem.php?stats=daily&msgid= tool>
I noticed that, on certain days, a lot of messages were still being
tagged by Nocem
So I'm going to take a look at the types of messages targeted by the
various nocems issuers.
I'll start with what surprised me most as a French speaker: there would
be daily spam on groups in the "fr" hierarchy. Having consulted a
sample of nocem, it's "nono le petit robot" who seems to produce them.
--------------------------------------------------------
"nono le petit robot"
Analysis(1) of March 31st nocems from 10:35am to 3:15pm
--------------------------------------------------------
Type "spam"
4 nocems over the period studied: <uubao6$qqd$1@rasp.pasdenom.info>, <uubei3$2dv$1@rasp.pasdenom.info>, <uubnbb$jnh$1@rasp.pasdenom.info> and <uubnkm$kd1$1@rasp.pasdenom.info>
That's 8 users messages, all on the "fr.soc.politique" group.
None of these messages is spam.
The false positive rate is 100%.
Type "spam2"
5 nocems over the period studied: <uubgaa$5g0$1@rasp.pasdenom.info>, <uubl9m$fna$1@rasp.pasdenom.info>, <uublj3$g8q$1@rasp.pasdenom.info>, <uubmoj$ie0$1@rasp.pasdenom.info> and <uubn1v$j48$1@rasp.pasdenom.info>
That's 6 users messages, all on the "fr.soc.politique" group.
None of these messages is spam.
The false positive rate is 100%.
Type "spam3":
2 nocems over the period studied: <uubelr$2k5$1@rasp.pasdenom.info> and <uublmq$gj7$1@rasp.pasdenom.info>
This represents 13 users messages, 11 of which concern the "fr.soc.politique" group and 1 message is not accessible (on nemo.bavardages),
the others are not spam.
The false positive rate is therefore 92%.
Conclusion: The spam detection success rates claimed by the robot(2)
are not reflected in the annalyzed sample. Instead of the 99.99%
success rate expected for the "spam" type, here we're at 0%.
The rest of the nocems review in a future post.
(1) Why didn't you analyze the whole day or a week or a month? I didn't
have the courage to go any further, given that the analysis is the same
for each nocem looked at and that it corresponds to the reading
experience of the "fr" groups where spam is not visible.
(2)  <https://pasdenom.info/nocem.html>
--
Admin of news.usenet.ovh