If you don't use any special characters, you could cut off the high bit of each letter, that would reduce the size 12.5%. If its all upper-case, you can do a direct mapping with only 6 bits which could compress it by 25%. Otherwise, huffman is really good at text compression (You might be able to get up to 50% compression) but the algorithm is so long and complicated that it might only be worth it if you had over 5k of text. Another suggestion is to use a special character that you never use as a "word replacer" for common words. Lets say that character [FB] is never used, you could have the sequence [FB01] replace the word "the" and [FB02] replace the word "only" etc. They can even be inserted into the middle of a sentence: [FB02]"re" would become "there".
