completely useless

[BloodKillerDeathMan]

Let me start by saying that I would love for someone to tell me why I
am wrong. Here is why I think this tool is useless:

1. No actual y-axis scale
1b. This means I can't search two separate sets of terms in separate
windows and have any hope of comparing the two either visually or
computationally
1c. This means a lot of interesting mashup applications are out

2. The "or" feature doesn't seem to work properly
2b. Try searching britney spears, paris hilton, britney spears |
paris hilton and you will see that people who supposedly searched for
either britney spears or paris hilton lie far beneath the graphs for
individual searches for either
2c. Try search britney spears, paris hilton on one browser window and
britney spears | paris hilton on another and again they do not seem to
match up. Of course, one cannot tell since the only clue about y-axis
is that the scale is linear. See 1c.

3. The regions, cities, and languages breakdown also include no
numbers
3a. I can see approximately whether mexico searched more for britney
or paris, but only by eyeballing it, which is not likely to be
accurate on that fine of a bar graph. How am I supposed to match this
data up with stock prices of clothing lines?

I have other rants about the tool, but I'll hold them, because the
ones listed above are all that matter, and here is why:

Alpha: I cannot reference this data ever because I cannot even
describe the numbers behind them. Showing any of these graphics is to
put myself in a completely indefensible position. Further, I would
argue that showing these kinds of "statistics" without the actual data
is irresponsible, and anyone trying to do so would get eaten alive by
their peers.

Beta: Throwing the previous reason, Alpha, out the window and
assuming that this could still be interesting from a completely
subjective, social, or other kind of view, I cannot script interesting
web services against it without parsing a bunch of html parsing and
string mining. I learned a long time ago not to code those types of
brittle applications.

Again, I would love for someone to prove me wrong.

[informatix]

It's very easy to prove you are in wrong
You say that google trend are useless ?!?
I'm sure you will go to use them very soon !!!

On Oct 24, 4:35 am, BloodKillerDeathMan

[Adam Kilgarriff]

Of course you're right - applciations like this are like promotional
samples of shampoo - just there to tiockle and entice. Not usable for
proper science. I wrote this up (from a language technology
perspective) in 2007
"Googleology is Bad Science" - http://www.kilgarriff.co.uk/publications.htm

On Oct 24, 3:35 am, BloodKillerDeathMan

[Chef Quickie Mister]

Agreed, except for the hyperbolic "completely useless." It is
interesting (at least for me) to see the amount of interest in a
subject in reference to itself. So, with me entertained, it has served
a purpose. Also, if anyone is irresponsible in the operation of this
tool it is the person who takes as fact the undefined approximations
it presents.
That said,
Why not define the Y Axis? Why not put some damn numbers on the page?
How about a zoom feature? Perhaps as my mouse rolled over the graph I
could see the exact date and number of searches defined by that pixel,
maybe I could see those definitions as I traced the graph.

Also, since the news results presented are often woefully irrelevant,
perhaps people could tag dates with descriptions of and links to
events that relate to a sudden spike or drop.

At the moment this is pretty damn useless, but with more information
involved it could prove to be wildly educational.

[magicwuh]

Agree... that this project needs improvement, but I think that
even now it provides good informations about a trend (!). which
means that users probably don`t want to see a high amount of
numbers, so lets keep it simple.
I searched for the trend of two different companies and I can see
on the first sight which company was searched more.
But I also think that additional Informations about the Y- Axis
(to see what means 'more') and more details of dates (I like your
idea of the zooming) would give additional Informations to users,
so he can decide weather it is useful for him or not.
Another idea:
Why not search for trends for a specific date?
lets take american idol for example. Perhaps it is interesting to
compare
all candidate searches after a show, or something like that.
I think a one month period is too unprecise for some needs.

[Lorenzo_]

I don't know if this can help, but try use the "or" this way:

britney spears, paris hilton, "britney spears" | "paris hilton"

Lorenzo