[mythtv-users] commercial detection always cuts off end of Colbert Report

Brad Templeton brad+myth at templetons.com
Fri Aug 18 20:40:40 UTC 2006


On Thu, Aug 17, 2006 at 02:57:15PM -0500, Matt wrote:
> Hello,
> 
>    I just wanted to say that Colbert Report is now one of my most
> recorded programs on my mythbox after I've been using it for 2.5
> years.  I notice that every episode, after the final commercial break,
> the show comes back on anywhere from 5 seconds to 60 seconds; but
> rarely does it EVER show that part of the show, it always gets lumped
> in with the commercials and always gets skipped (every time).  Anyway
> that the commercial detector can be optimized or fixed so this doesn't
> happen?
> 
> Thanks!
> Matt

I continue to think that the right answer to these and many other
special case problems is not cleverer general heuristics, but
a collaboration system that allows individuals to come up with
specific tricks for specific shows and publish those tricks out
to the community.

By which I mean a very simple language to describe things like
commercial skip tricks (and also as discussed earlier actual notes
about the real location of commercials, as determined by watching
how people FF/Rew and otherwise correct the actions of the
default commercial detector,) special padding rules (such as that
clever new system that queries the web to find out if a sporting
event has actually finished) and more.

The key is a means so that if one person solves a problem, it's
solved for all of us.   Of course, that's what an open source
project is when it comes to code contributions.  

But I speak of something halfway between code contribution (which
normally would not involve special coding for the Colbert Report)
and listings data (which is nothing but data about specific shows).

A way to say things like:
    a) This show has a snippet of programming at the end
    b) This show commonly starts a little early
    c) The last N seconds of this show are commercials or unimportant
        credits.
    d) The first N seconds of this show are important/unimportant
    e) Commercial break expectations.  For example, you can say that
       60 minutes always has a leader, a commercial break of 2-3
       minutes, a 12-17 minute segment, a commercial break of 2-3
       minutes, a 12-17 minute segment, a commercial break of 3-5
       minutes, a 12-17 minute segment, a commercial break of 2-3
       minutes, a 2-5 minute segment and then commercials until end.

       A clever commercial analysis program could use lists like this
      (written as simple expressions) to do highly accurate commercial
      handling.

    f) This URL will have live information on when the show ends.
       (Could be sports sites, or just a site that waits for reliable
        reports from live watchers of things like the Oscars)
    g) Corrections to standing errors in listings data, or human
       detected errors in the listings data.
    h) Repeat detection, after the fact, or sometimes before, when
       the listings don't tell you. 
    i) Ratings of really good shows by people who saw them soon
       on an early broadcast, suggesting you record them on later
       broadcasts.   Example -- suddenly everybody is talking about
       a new show, you wished you had recorded it.

Of course, if I had time to code this, I would be doing more than
blow smoke here.    Each of these ideas above requires some
serious work.  However, they all need, underneath, a mechanism to
collect and broadcast such information.    

Some of it would be real time which is hard.  Some of it is "per season"
or "per show" and just requires one reliable person to write the rules
for handling that show, and another reliable person to vet that the
rules are valid and not trying to screw with things.

I have done some of this.  Tvwish allows you to put show record
recommendations on a web page and others can import them, but I don't
know if anybody's doing it yet.


More information about the mythtv-users mailing list