Frequencies of occurrence
---------------------------
Part of his doctoral thesis, published in 1952, was devoted to information theory and, more specifically, to the frequency of words in a given text.
This is a very old question: the earliest work on the subject dates back to the 1380s, beginning with Cicco Simonetta’s Liber Zifrorum. At the time, cryptographers compiled frequency tables for letters, bigrams (sequences of two consecutive letters) and trigrams in a given language. Here is the beginning of the letter-frequency table for contemporary French:
| 12.1% | 7.11% | 6.59% | 6.51% | 6.39% | 6.07% | 5.92% | 5.02% |
|---|---|---|---|---|---|---|---|
| e | a | i | s | n | r | t | o |
More recently, the French stenographer Jean-Baptiste Estoup (1868–1950), one of the founders of statistical methods for analysing texts, took an interest in the problem. As Mandelbrot notes: "While studying the problem of stenography, J.-B. Estoup observed an apparently hyperbolic relationship between the number of new and distinct words in consecutive samples of 1,000 French words, on the one hand, and the cumulative word count, on the other."
