Gawk -- 336 (originally 507) characters
(after fixing the output formatting; fixing the contractions thing; tweaking; tweaking again; removing a wholly unnecessary sorting step; tweaking yet again; and again (oops this one broke the formatting); tweak some more; taking up Matt's challenge I desperately tweak so more; found another place to save a few, but gave two back to fix the bar length bug)
Heh heh! I am momentarily ahead of [Matt's JavaScript][1] solutioncounter challenge! ;) and [AKX's python][2].
The problem seems to call out for a language that implements native associative arrays, so of course I've chosen one with a horribly deficient set of operators on them. In particular, you cannot control the order in which awk offers up the elements of a hash map, so I repeatedly scan the whole map to find the currently most numerous item, print it and delete it from the array.
It is all terribly inefficient, with all the golfifcations I've made it has gotten to be pretty awful, as well.
Minified:
{gsub("[^a-zA-Z]"," ");for(;NF;NF--)a[tolower($NF)]++}
END{split("the and of to a i it in or is",b," ");
for(w in b)delete a[b[w]];d=1;for(w in a){e=a[w]/(78-length(w));if(e>d)d=e}
for(i=22;i;--i){e=0;for(w in a)if(a[w]>e)e=a[x=w];l=a[x]/d-2;
t=sprintf(sprintf("%%%dc",l)," ");gsub(" ","_",t);if(i==22)print" "t;
print"|"t"| "x;delete a[x]}}
line breaks for clarity only: they are not necessary and should not be counted.
Output:
$ gawk -f wordfreq.awk.min < 11.txt
_________________________________________________________________________
|_________________________________________________________________________| she
|_______________________________________________________________| you
|____________________________________________________________| said
|____________________________________________________| alice
|______________________________________________| was
|__________________________________________| that
|___________________________________| as
|_______________________________| her
|____________________________| with
|____________________________| at
|___________________________| s
|___________________________| t
|_________________________| on
|_________________________| all
|______________________| this
|______________________| for
|______________________| had
|_____________________| but
|____________________| be
|____________________| not
|___________________| they
|__________________| so
$ sed 's/you/superlongstring/gI' 11.txt | gawk -f wordfreq.awk.min
______________________________________________________________________
|______________________________________________________________________| she
|_____________________________________________________________| superlongstring
|__________________________________________________________| said
|__________________________________________________| alice
|____________________________________________| was
|_________________________________________| that
|_________________________________| as
|______________________________| her
|___________________________| with
|___________________________| at
|__________________________| s
|__________________________| t
|________________________| on
|________________________| all
|_____________________| this
|_____________________| for
|_____________________| had
|____________________| but
|___________________| be
|___________________| not
|__________________| they
|_________________| so
Readable; 633 characters (originally 949):
{
gsub("[^a-zA-Z]"," ");
for(;NF;NF--)
a[tolower($NF)]++
}
END{
# remove "short" words
split("the and of to a i it in or is",b," ");
for (w in b)
delete a[b[w]];
# Find the bar ratio
d=1;
for (w in a) {
e=a[w]/(78-length(w));
if (e>d)
d=e
}
# Print the entries highest count first
for (i=22; i; --i){
# find the highest count
e=0;
for (w in a)
if (a[w]>e)
e=a[x=w];
# Print the bar
l=a[x]/d-2;
# make a string of "_" the right length
t=sprintf(sprintf("%%%dc",l)," ");
gsub(" ","_",t);
if (i==22) print" "t;
print"|"t"| "x;
delete a[x]
}
}
