The Phonetic Information Redundancy in Mandarin: A Support for Information Optimizing Model of Language

摘要

This paper introduces a language model referred to as Information Optimizing Model (IO Model), which proposes that language is an optimal encoding system to communicate in the most efficient way, based on following rules of language: Efficiency, Cooperation, Distinction, Evolution, and Reliability. In China, when confusable phonetic contrast (PC) such as [s/sh] is used in Mandarin communication between Beijing Dialect speaker and Southern Dialect speaker, they have to guess what the other want to pronounce. Phonetic statistics done for Mandarin and its analyses show that when Maximizing strategy is adopted and tone information is used, guess errors of confusable Initial PC [s/sh], [c/ch], [z/zh], and [l] are as low as 6.5~12.7%, and guess errors of confusable Final PC [en/eng], [in/ing], and [an/ang] are 10.9~24.1%. Entropy calculated for Mandarin syllables also shows that the tone information added to syllables not only increases the entropy of syllables by 1.05 bits, but also increases the mutual information between Initials and Finals by 0.77 bit. These are evidences of existence of phonetic information redundancy in language, and support Reliability rule of IO Model we have proposed.