I've been working on splitting the decomposable words into separate
morphemes.
So far there are 1548 out of 9341 words that have a morpheme "-s" on the
end. That's over 16% redundancy with just one morpheme that's already in
your list. I have already identified about 1548 (16%) roots, 4599 (49%)
known words with multiple morphemes, and 3194 (34%) words still to check. So
I estimate that well over half of all the word entries are redundant.

I have also discovered a new "morpheme": devoicing the final consonant of a
verb to make a noun indicating a single occurrence of the action of the verb
(for which I'm sure there is a more succinct term ('instantial'?)).
I have identified the following words in the list that have this "morpheme":
advise/advice
believe/belief
breathe/breath
choose/choice
lose/loss
There are some vowel changes going on here too that I haven't tried to
figure out yet.
And I suppose it's possible to consider the morpheme to be one of voicing
the final consonant of the 'instantial' noun to form the verb.

stevo

On Wed, Jul 21, 2010 at 7:46 PM, Gary Shannon <[log in to unmask]> wrote:

> Good point. I've been thinking what I might do is squeeze out all the
> redundancy that results from inflections, make the inflections
> separate codes, then encode the whole thing using a REAL Huffman tree,
> BUT using alternating vowels and consonants so that the resulting
> words would all be pronounceable.
>
> Then it would be a fairly complete lexicon, with word length optimized
> for word frequency, pronounceable, and no longer tied to English
> inflection rules. In other words, a usable conlang lexicon.
>
> --gary
>
>
> On Wed, Jul 21, 2010 at 2:55 PM, MorphemeAddict <[log in to unmask]> wrote:
> > There's a huge amount of redundancy in that list of words. E.g., there
> are
> > about 565 words that end in a suffix "-s".
> >
> > stevo
> >
> > On Wed, Jul 21, 2010 at 3:00 PM, Gary Shannon <[log in to unmask]> wrote:
> >
> >> I tried that with a spreadsheet to compute how many distinct codes I
> >> would get using various ways of dividing up the pie. The problem I
> >> found with one-letter codes is that they ended up severely reducing
> >> the number of available two and three-letter codes. Using those six
> >> one-letter codes eliminates 120 two-letter codes, 720 three-letter
> >> codes and  18,720 four-letter codes. In the end, plugging numbers into
> >> the spreadsheet, sticking with two-letter minimums gave me the best
> >> result.
> >>
> >> There is most certainly a more optimum way to do it, possibly even
> >> using some one-letter codes, but just dinking around by hand with a
> >> spreadsheet, that's the best I could come up with. Then again, I was
> >> more trying to make sure I had enough codes to go around and less
> >> concerned with the compression ratio. Making compression a priority
> >> would certainly result in a different set of assignments.
> >>
> >> --gary
> >>
> >> On Wed, Jul 21, 2010 at 11:42 AM, Philip Newton <
> [log in to unmask]>
> >> wrote:
> >>
> >> >> http://fiziwig.com/crypto/huffcode.html
> >> >
> >> > I wonder whether you would have gained anything by using the 6
> >> > one-letter codes "T, U, V, W, X, Y, Z" for the six most common words
> >> > of the English language. (Or actually, six most common morphemes,
> >> > since you also encode certain combining morphemes such as dis- and
> >> > -s.)
> >> >
> >> > Cheers,
> >> > Philip
> >> > --
> >> > Philip Newton <[log in to unmask]>
> >> >
> >>
> >
>