It's important to put a date range into the queries.

You might find that "Amnesty" was more in the news during 2005, and NATO during 2006. A site whose archives cover both years might rate equally for both terms, yet a one-year archive of the same site would suggest a bias one way or the other.

does not make sense.. complete aggreation of the term shows me this.. and whatever happened after 2005 eh ?? "Bush is" dissappers ??

http://slash.pd.googlepages.com/saidvsis.png

True Roger, archive scope is definitely a part which skews the results even more for this approach!

site:foxnews.com "bush misled" = 40
site:nytimes.com "bush misled" = 493

liberal bias?

I don't get anything like the ratios you report for bush AND success vs bush AND defeat on the two sites you mention.

The actual ratio in favor of success at foxnews.com is, by your own numbers, 10 to 1, not 9 to 10 as you state. The ratio in favor of success at alternet.org is only 2 to 1 not 4.6 to 10 as you state.

Alternet.org has a lot more hits for either phrase than fox – I guess because foxnews.com is all about video?

site success defeat ratio
alternet.org 203,000 109,000 1.9
foxnews.com 10,300 948 10.9

Philipp does not mean '9 to 10' he means 9/10, 90%.

IE that for every 10 articles about containing the words "Bush" AND ("success" OR "failure"), 9 contain "success" and 1 contains "failure".

You might want to rephrase it Philipp.

At first Fox News looks not as biased as it's repution suggests:

site:foxnews.com bush "war criminal" – 68

They have really critical reports like this one:
http://www.foxnews.com/story/0,2933,194248,00.html

But then again if you compare: site:alternet.org bush "war criminal" – 34.100

Yeah, I meant "4.6/ 10" not "4.6 to 10" in the first paragraph. I fixed this.

Why don´t you try: Bush +success in news.google.com and news.yahoo.com?
Google is pro Bush by far

Note it's about the relative proportion of Bush + success vs Bush + defeat in comparison of at least two news sources, in other words you need at least 4 Google queries to make this work. Everything else will just turn up mere quantities, which aren't revealing (e.g. a site may simply be bigger, or on average the word "success" [or "defeat"] may simply be too common).

"site:foxnews.com "bush misled" = 40
site:nytimes.com "bush misled" = 493
liberal bias?"

Or conservative brush under the rug?

Either way, these kind of text searches have absolutely no utility other than sparking partisan rants.

As much as I'd like to say this is yet another sign of "proof" of how left/right leaning news organizations are (and Philipp picked two prime examples IMHO), if you "normalize" the results based on site size, the results show they don't really mean much.

site:foxnews.com "bush misled" = 40
site:nytimes.com "bush misled" = 493

site:foxnews.com bush=175,000
site:nytimes.com bush=15,100,000

Conclusion: nytimes.com is a LOT bigger than FoxNews.com ... and looking at the ratio's, bush "misled" a lower percentage on the NYTimes ... even though most people would argue that GWB is NOT a favorite of the NYT.

>whatever happened after 2005 eh ?? "Bush is" dissappers ??<
Not enough searches for this query.

Alek, agreed, you need to pit ratios against each other, not count quantity – so the example ...

<<"site:foxnews.com "bush misled" = 40
site:nytimes.com "bush misled" = 493>>

... just doesn't work. You need 4 queries to make 1 comparison, and you can only compare ratios then, not absolute quantities. As you point out...

Just remember, "There are three kinds of lies: Lies, Damn Lies, and Statistics." – Benjamin Disraeli