Article,
Mechanical Translation and Zipf's Law
Language, 33 (4): 545-552 (1957)
Abstract
A problem which arises in the course of research on mechanical translation is the prediction of dictionary size. This article investigates the relation between empirical frequency laws and the function V(n)-the expected number of different words in an n-word sample of text. It is found that the probability-law proposed by Joos (1936) yields results which do not check well with experiments, and it is concluded that some modification of it is necessary for the purpose of vocabulary prediction.
