Statistical properties of different languages based on entropy and information theory

ABSTRACT Master of Science Thesis STATISTICAL PROPERTIES OF DIFFERENT LANGUAGES BASED ON ENTROPY AND INFORMATION THEORY ŞENAY YOLAÇAN Anadolu University Graduate School of Sciences Statistics Program Supervisor: Prof. Dr. Aladdin ŞAMİLOV 2005, 193 pages In this thesis, statistical, properties, which are letter frequency distributions, entropy per letter and information measure per symbol, of different languages as Turkish, English, German, French, Russian and Spanish were examined. Utilizing these statistical properties, binary codes for these languages were constructed by Ordinary Coding Method, Shannon Coding Method, Shannon-Fano Coding Method, Improved Fano Coding Method and Shannon-Fano-Elias Coding method and results were established as tables. Coding method that realize optimal coding was determined to be Huffman Coding Method. Moreover, each of different languages taken in to account was considered as a coding type. In order to determine optimal language in the sense of coding theory, statistical properties of translated papers on the same semantic content in these languages were examined. In this investigation, by utilizing these statistical properties, it was established that English language uses least symbol than the others to express the same semantic content. Then Turkish language was determined as second language after English language. Keywords: Entropy, Information, Coding, Optimallity, Turkish u