Multiple pronunciation model for Amharic speech recognition system

In this paper the research have tried to show the pattern variations of sound units in Amharic language for multiple pronunciation model. This are variation of sound units at lexical level due to dialects. After that an attempt to build a pronunciation dictionary for Automatic Speech Recognition (ASR).At last comments and recommendations are included. Amharic is an official language of Ethiopia. It is a Semitic language that has the greatest number of speakers after Arabic. Amharic has five dialectical variations spoken named as: Addis Ababa, Gojam, Gonder ,Wollo and Menz[1]. The Amharic writing system uses multitudes of ways to denote compound words and there is no agreed upon spelling standard for compounds. As a result of this and of the size of the country leading to vast dialectal dispersion, lexical variation and homophony is very common [2]. Pronunciation variation is a phenomenon observed within a speaker or within a group of speakers of the same dialect or among speakers across dialects of the same language. Pronunciation variation deals with the different ways of speaking a given word. Pronunciation variation modeling has been studied in the field of speech synthesis and recognition to improve performance of the corresponding speech systems [3]. The Amharic orthography as it is represented in the Amharic character set consists of 276 distinct symbols. These symbols are classified into four groups. In the first category (33*7=231) there are thirtythree core orthographic symbols, each of which has seven different shapes, usually known as orders, to represent the seven vowels. Each consonant and the seven vowels in combination represent CV syllables[4]. Each of these consonant and vowel grapheme can appear independently or can form a combinant letter. Each consonant can form CV pattern except with the vowel /ix/ (called epenthesis vowel)[5] . The second category (4*5=20) consists of four labio-velar symbols, which have five orders. The eighteen labelized consonant, which have only one order, are the third category. The fourth category is the representation of numbers from 1 to 10 and multiples of 10 each with different symbols [4]. The Amharic language script is called Ethiopic. Even though the vowel modification is not entirely semantic, the Ethiopic script is a syllabic structure [5]. The first International Workshop on Spoken Languages Technologies for Under-resourced languages (SLTU 2008) The first International Workshop on Spoken Languages Technologies for Under-resourced languages (SLTU 2008)