[mythtv-users] Daddy, what's a commercial

Jared C. Davis jared at cs.utexas.edu
Sat Jan 13 16:13:31 UTC 2007

On 1/12/07, Tim Litwiller <timlitw at gmail.com> wrote:

>  Hmm. I suppose we could route these questionable caption blocks thru
> spamassassin and use the bayes filter to help determine, based on the use of
> words, the "score" of that block of video.

That sounds like a wonderful idea.  Commercials just aren't written to
avoid this kind of filtering.

Probably half of the commercials on the evening news are for new
drugs, and almost all of them have common phrases like "ask your
doctor", "common side effects include", "if you're pregnant or
nursing", "is not for everyone", etc.  It would probably be trivial to
filter these out based on the caption text.  Similarly, investment
commercials that often air during political shows often will say "may
lose value", "consult your financial adviser", etc.  I don't really
watch childrens' shows, but perhaps their text would have similar

And beyond that, almost every company says at least their name and has
some slogan they'll say in every commercial: "built ford tough",
"always coca cola", "this bud's for you", "the independent spirit",
etc, which would be easy to catch.

The one danger I see is shows where they integrate the advertising,
particularly sports shows.  When a football game or tennis match comes
back from commercials, they'll often say who they're sponsored by, and
mention the slogan of that company.  This might fool the filter into
flagging part of the program, but looking for 30 second blocks might
alleviate this.  Also, you wouldn't have to just have negative
filters.  In a football game you'll hear the announcers saying "first
down", "touchdown", "to the forty, the thirty, the twenty", etc.  So,
seeing phrases like these could boost a block's "good words" score.
Similarly, in the news, you'll often hear "Tonight we're going to take
a closer look..." as a common transition.

Apparently there are some regulations about most programs needing to
have closed captions (at least in the US):

I noticed that promotional announcements under 10 minutes in duration
are exempt, but if commercials started doing that, you might be able
to just start using the absence of captions as a method.

I should have prefaced all this by saying I have no idea how MythTV
works and perhaps it already does this kind of thing, so I may have
inadvertently suggested something that is already implemented.
Commercial skipping works pretty well for me already, but I do hate
commercials... :)

Jared C. Davis <jared at cs.utexas.edu>
3600 Greystone Drive #604
Austin, TX 78731

More information about the mythtv-users mailing list