Extracting information from a keyboard…

Yesterday, Baptiste published a post on “ethno-photography” (here). As he mentioned it, in Paris 8, they experience a real absence of serious cleaning of office equipment. He then shows the keyboard of the only computer they can use in the sociology department (for forty researchers),

Apart from the fact that everyone in France should be ashamed to see how much is spent in universities (which is the first information we have from that picture), we should also be able to guess in which langage people work in this department.
I considered three books (two in French, one in English) and I would like to see the frequency of each letter,

  • Mauss, manuel d’éthnographie (here), 1926
  • Durkheim, Livre II: Les croyances élémentaires in Les formes élémentaires de la vie religieuse (here), 1912
  • Ferri, Criminal Sociology (here), 1896

Those three books are in rich text format, I just changed it to get text files… Then, it is easy to count appearance of letters. E.g. for Mauss,

> library(corpora)
> textfile=scan("MAUSS-manuel.txt",
+ what="char", sep="\n")
Read 1550 items
> textfile<-tolower(textfile)
> M=NA
> for(i in 1:length(textfile)){
+ line=textfile[i]
+ M=c(M,strsplit(as.character(line),"")[[1]])
+ }
> T=table(M)
> T
    '     -           \t     !     "     %     &     (     )     ,     .     / 
 5308  1049 86589    44     3     3     2     2   370   391  6609  4909    12 
    :     ;     ?     @     ]     _     ~     ’     =     «     »     ¬     ° 
  819  1178   113     1     1     4     1    39     1   108   107   823     3 
    …     0     1     2     3     4     5     6     7     8     9     a     à 
    1    69   213    83    73    34    48    33    28    64   151 30559  1651 
    â     ä     b     c     ç     d     e     é     è     ê     ë     f     g 
  224     3  3562 14678   110 17713 63955 10354  1798  1000     5  4555  4911 
    h     i     î     ï     j     k     l     m     n     ñ     o     º     ô 
 4359 30851   226    47  1147   247 24792 12844 32525     6 25562     2   151 
    ö     œ     p     q     r     s     t     u     ù     û     ü     v     w 
   12    52 12696  4667 28237 37630 32945 25001   211    40     9  4787   164 
    x     y     z 
 1996  1222   343

Then, we can summarize in to see proportion of standard 26 letters, and we have, for Mauss,

and for Durkeim,

If we compare the two, we have almost the same proportions,

If we look at our book in English now, we have

i.e., if we compare with Mauss for instance

So we have much more E in French than in English, but still, people writing in English use a lot the E. So looking at the E should not give us any clue…. But we can see that in English, the H is as common as the L, or the C. Not in French, where L is much more frequent than the H. But on the picture, the C is more clear than the H. We can also look at the U, which is common in French, not in English… Here, on the keyboard, it is perfectly clear… so I guess people use it frequently.
So I would say that they write more in French than the write in English, on that computer.
Actually, the same idea has been used a long time ago on calculators to see that Benford’s law works: some numbers are really used (as well as the legend pretends that some pages in logarithm books were never used….), see here orthere. So Baptiste, if one day the keyboard is cleaned up, please send me another picture after a few weeks to see if things have changed….
An for those who cannot imagine how it is to work in some universities in France, just look at his blog (here). Pictures are unbelievable….Good luck Baptiste….

3 thoughts on “Extracting information from a keyboard…”

  1. And how about quantifying the accurate levels of dust to have a clearer picture? Did it for light intensity on my worn-out Dell keyboard, if you have any comments on the method or the results (sorry it’s in French): http://gambette.blogspot.com/2008/1

    ANSWER:thanks Philippe… I did not know that ! it is great and extremely interesting… a lot of interesting information indeed… I observed that my results are not far away from those obtained here and there. Thanks again for the link !

  2. Hi Arthur,
    In French litterature, traditionnaly the printers consider the most common letters to be ESANTIRULO. Your code just proves that this belief is true…

    Moreover, in French, bold text is called “gras”, meaning “fat” – because of the quantity of ink needed to print a bold text. your pictures proves this to be right as well for keyboards !

    ANSWER: thanks Emmanuel for all these informations… and glad to see that that quick code was not bad..

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.