pageoneresults - 12:45 pm on Feb 19, 2011 (gmt 0)
But any machine algo is going to make mistakes some percentage of the time.
Yes it will. And, you have some level of control over it. Machine algo? Well, the machine algo needs machine readable grammar to function at its best.
And so far, they have given us no way to participate in this part of experiment.
Oh yes they have! Proper HTML formatting, semantic structure, etc. Most of the title edits I've seen have come from <h> elements on the page and/or that which immediately follows the <h> element. If no <h> element is there where they are expecting to see it, then the next best thing is used, whatever is there, in this case, a <p> element where it appears an <h> element should be?
I've seen more positive results with valid HTML and use of proper markup. Fragment IDs on <h> elements appear to also be a factor in these title edits.
A machine algo needs machine readable grammar. Its best if that machine readable grammar is without errors too. You know, that whole validation thing that everyone blows off.
Based on my experience, many of those title edits come from markup that immediately follows the <body> element. I don't see titles getting generated for stuff that sits further down in the HTML. But, I don't look at this stuff every day like you folks do either. :)
P.S. I wonder why this doesn't happen to the documents I manage? Hmmm, must be the valid, semantically correct HTML that does it.