The measurement of comentropy transfer rates
pp. 319-330
in: , Semiotics 1980, Berlin, Springer, 1982Abstract
Most people have heard of a game that children play called "Hangman." Each player tries to guess serially at the letters in the word that the other player has chosen, and every letter guessed incorrectly contributes to a penalty. But most of the time, the letters are correctly guessed. Given a good familiarity with a language, say American, most people would guess that the letter "n" would usually follow, if "a" were the preceding letter. Not only that, but the more letters that we know in a word, the easier it becomes to guess at the succeeding letters. Thus, if we are shown the following, "Experi," most people would easily guess that what followed would be "ment," since there are only a few logical choices. So it seems that there is a lot of redundancy built into a language such as American, or for that matter, any other language. But the problem remains in how to obtain a good measurement of this level of redundancy. How can it be measured?