Using a large synchronous Chinese corpus, we show how word and character entropy variations exhibit interesting differences in terms of time and space for different Chinese speech communities. We find that word entropy values are affected by the quality of the segmentation process. We also note that word entropies can be affected by proper nouns, which is the most volatile segment of the stable lexicon of the language. Our word and character entropy results provide interesting comparison with the earlier results and the first-ever average joint character entropies (a.k.a. entropy rates) of Chinese up to order 20 provided by us indicate that the limits of the conditional character entropies of Chinese for the different speech communities should be about 1 (or less). This invites questions on whether early convergence of character entropies would also entail word entropy convergence.
展开▼
