The not-so-slow death of truthiness?

[Clai Rice]

Accuracy in Truthimetrics

Thanks to Lynne Murphy and Ben Zimmer, who pointed me to discussions of
weaknesses in Google search numbers. Ben had argued that Google was not
to be trusted for hit counts over 1000
(http://itre.cis.upenn.edu/~myl/languagelog/archives/002658.html). I
would like to add that they aren't reliable even when considering low
numbers, lacking what Ben called "internal consistency" when comparing
hit counts among Google results alone. My results for 'truthiness
-colbert' on Google Groups follow Dave's roughly, except that I get all
zeros for the months in 2005:

Dave's:         Mine:
>
> 2005
> Oct.    6             0
> Nov.    1             0
> Dec.    0             0
>
> 2006
> Jan.    41    41
> Feb.    77    189
> Mar.    45    45
> Apr.    43    33
> May     42    47
> Jun.    54    43
> Jul.    38    36
> Aug.    68    30 (through 17 Aug.)

My results are given without "repeating the search with the omitted
results included." But what's up with Feb? My exact search string is:
http://groups.google.com/groups?as_q=truthiness&num=50&scoring=d&hl=en&a
s_epq=&as_oq=&as_eq=colbert&as_ugroup=&as_usubject=&as_uauthors=&lr=&as_
qdr=&as_drrb=b&as_mind=1&as_minm=2&as_miny=2006&as_maxd=28&as_maxm=2&as_
maxy=2006&safe=off

While the kids were in the bath I went through January day by day:
Jan
01      1       11      4       21      9
02      0       12      6       22      7
03      0       13      6       23      8
04      0       14      4       24      14
05      0       15      3       25      11
06      0       16      4       26      12
07      7       17      12      27      7
08      7       18      9       28      8
09      3       19      14      29      5
10      3       20      10      30      12
                                31      3

The total here is 189, as compared to 41 when searching by the month.
When searching Jan by the month I can't get more than 50 hits by
changing any settings, omitting the -colbert limitation, or trying
misspellings. When searching dates 1-10-06 to 1-10-06, Google sometimes
returns posts dated in the early morning hours of 1-11. But I didn't see
any repeated hits.

If we take the bare distinction none vs. some, however, it still looks
like the WOTY vote shoved the word into public consciousness (all the
hits on Jan 7, for example, are reporting the WOTY decision).

Clai Rice

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org