You can define it in terms of language as such “whether the journalist uses rhetorical artifacts that distort the information to support his opinion, or not.” (1)
More definitions and discussions can be found in the paper.
Do you mean for the framing of a piece, or the decision on what to report and to which degree/frequency? The former is more simple -
1. HORRIBLE MASS GENOCIDE BEGINS IN SPRINGFIELD (partial, even though it's within todays overton window)
2. FINALLY, AWESOME MASS GENOCIDE BEGINS IN SPRINGFIELD (partial, outside the overton window)
3. MASS GENOCIDE BEGINS IN SPRINGFIELD (impartial reporting)
Headline is obviously far more simple though, removing most adjectives does most of the job, and goes a long way in the content itself as well. Something like choosing which photos to run or editing of video clips would be a lot harder to quantify impartiality/bias, both measuring and for the person doing it I think. As for overall article frequency and depth of article per subject, I think that's where your Overton window question becomes far more difficult - you could quantify a given piece or source against all other news sources and find the edges and the median, but yeah, in most cases just everything is biased by the window at the macro level. You could go back in time and find the dead center median bias media source publishing glowing reports on bloodletting and public executions.
