This is a great brainstorming tool for doing light blogspam-style "journalism", if your readers don't know about it (which they probably don't). Draw something with a rather unlikely looking feature— a graph that strikes the reader as quite definitely not just random noise. See if you get anything interesting in return. If yes, conjure up a backstory, and write up a post in a confident, assertive tone, that closes with the graph as your "care to explain this?!" evidence nailing home your suspicion.
In a way it's even an improvement on confirmation bias. With confirmation bias, you have a conclusion you want to sell, and selectively search for evidence that might support it. But with this suggestion, you search for interesting evidence, and then invent a conclusion that it might support, post-hoc.
It's not necessarily as bad as it seems because this is relative to overall search term popularity. As more non-techies adopted the Internet, the prevalence of all techie topics declined, some more than others. E.g. Python is more popular as ever, but it's search term interest peaked in 2006, even though it's absolute search numbers increased nicely.
I don't think google do trend graphs for absolute search numbers, so one way to approximate that is to compare the graphs for competing terms using Google Trends (1). So to get a feel for how popular Python is, compare it's trend graph to say Java or Ruby.
Another factor I suspect may be relevant for Linux is that the rise of prominent distros has probably sapped away search term activity for general searches for Linux.
foss is a nerd thing. The general consumer has no idea what source code is in the first place, they often don't know what an application or OS is, they just click the thing on the screen to make facebook appear.
So it loses search percentage because it isn't becoming a mainstream concept, which is sad. People should care more about their software freedom but they don't even know what software is.
The number of people, globally, using the internet, is still on that same near exponential growth curve that the first world nations were seeing in the late 90s. It really hasn't subsided; only gone to lower income brackets.
Heck, half of the US didn't have reliable access to the web until I think 2002?
And these late adopters, they are both non-english speakers and non-technical people.
Additionally, the early adopters, let's take my parents, who had Prodigy in about 1990, are now using the web more for things like shopping and content consumption, then they did in 1998.
If they wanted to watch the latest TV show, they can do that now online, AND they know this. Compare that to 2004. That will be reflected in the search terms.
Also, there are internet phenomenons that are at the order of that which we have not seen. Psy's videos on youtube for instance; 1 ba-ba-billion+? That's insane.
So yeah, Linux is left in the dust. Also, it's much easier than it used to be - take that from someone who's been using it since 1996 (?)
I need to do a write-up some time soon of how good we have it now - for my own memory.
I believe twitter didn't start seeing crazy growth until 2009 once they focus on activating their users and moving away from the idea that everyone needs to tweet to use twitter.
Just Google these words and add the word 'summer' to get some ideas. Here's what I came up with:
Bruising: More children play outside in the summer → more bruises → more parents/children(?) Googling for it.
Brown spots: Might be referring to brown spots in someone's lawn due to the weather? Or summer freckles?
That's a pretty legitimate question. At that time, LED TVs were starting to be marketed pretty heavily. I myself didn't understand what an LED TV was. Turns out it's an LCD but with an LED backlight, so it runs cooler and uses less electricity.
The question isn't really the difference between LED and LCD, it's the difference between an LCD TV and an LCD TV with an LED backlight.
The whitepaper describes an approximate nearest neighbor algorithm, but I thought this would use some form of signal decomposition...Discrete Cosine Transform or Fourier Transform.
"Please sign in to use this tool." That's all I see when I am not signed in. Too confident on Google's part that the people will always be signed in or do they actually need any personal info to use this tool?
Maybe it's a rate limiting thing? This is a relatively expensive search, and they probably don't want people to mechanically download all the graphs. The UI does support uploading your own time series data.
