10-Letter English

Careful observation of English text reveals that there are a few letters that could be safely removed from the writing system of English without much loss in legibility. For example, we could always replace x with ks, put kw in place of qu, and put dzh or something instead of j.

To most people, this would seem like a stupid or ridiculous idea. Why get rid of perfectly good letters and spellings just to be fractionally more efficient? But to the kind of linguist that I am, an interesting question arises: How few letters can one use to approximate English words closely enough that your text is translable into English, without having each letter be stood for by a series of letters? What if each letter can be represented by a series of letters of any length?

The second question is almost trivial if you know about ASCII. You can represent each letter with a specific five-letter combination of only two distinct letters with no loss at all, or with a four-letter combination if you push it to condense.

The first question requires much more consideration. Can we group similar letters into one symbol so that someone reading the mutated English can figure out what words mean (based on context and sound) with reasonable efficiency? Predictably, the answer is yes, but how low it can go is not obvious at all. This was my first attempt:

a d e g i l n p s u
a
q		 d
t		 e
o		 c
g
k
x		 i
j
y		 l
r		 h
m
n		 b
f
p		 s
z		 u
v
w

See if you can decipher this: "Dne leasen in indelesded in dnis sisden is pegause id seens de naue eneugn leddels de pe iusd paleli legiple unile ad dne sane dine nauing dne leeg ep genplede and uddel gippelisn." It's not all that bad, except for a few context errors. There's one problem, though: qu looks a lot like av, which really hurts a decription. Later, I put the q in with the g's.

Yes, I know that the "a" only has one letter. You can go lower, as with the 8-letter idea below, but at that point you are really sacrificing legibility, and text becomes inhibitively difficult to read.

a d e g l n p s
a
o
u
w		 d
t		 e
i
j
y		 c
g
k
q
x		 l
r		 h
m
n		 b
f
p
v		 s
z

The final development in this system was my recent idea of using numbers for the letters. I chose numbers that looked like the letters, but it isn't quite perfect. You can use the last number for a space if you want (you DO need spaces in this) or you could just replace the a back in. Either way... it's OK.

0 1 2 3 4 5 6 7 8 9
e i s n u l p d g -
a
e
o		 i
j
y		 s
z		 h
m
n		 u
v
w		 l
r		 b
f
p		 d
t		 c
g
k
q
x		 -

I'm thinking of using this for a code in games or something, so when you enter in your password it's just a number. :) That will be fun. The password is a word but you don't have any letters! Anyway, that thing back there translated into this would be

7309500203913913705027079139731292127039129608042091792003297
0930409030483950770529709609142796050519508165094315090797309
203097130930413897309500890698036507090379477059816605123.