Language differentiation based on sound patterns of the spoken word

Abstract

A categorical analysis was made of five languages. The sounds of speech were simulated using written text converted via International Phonetic Alphabet (IPA). The sounds of speech were identified as members of fricative, nasal, stop, or vowel categories. A statistical analysis was performed on categorical content of one (at various positions in the word), two, and three sound combinations. Several attempts to achieve a differentiation scheme were made before any success was realized. Two methods of developing conditional expectation are compared: Bayes' Conditional Probability Rule, and Cook's Prognostic Progression. Statistical analysis and "loop" tests indicated that languages do have unique patterns and can be differentiated on the statistics made in the first three sounds. !00% correct decisions were achieved for as few as five words in the loop test. Limited data base negated result significance beyond three successive sounds.