Fonetikai tulajdonságok alapú ábécé készítése

Abstract

The numerical representation of the letters in text processing and natural language processing task are usually based on the ordinary alphabet. This alphabet omits the phonetic features of the words, however these features has effect on the grammar. There is no distance defined in the traditional alphabet – the position of the letters is independent from the phonetic features. The proposed representation in vector space is based on the phonetic characteristics of the letters. The dimension reduction of the vector space into a one dimensional subspace yields an ordering of the letters which is based on phonetic features. The yielded alphabet has been shown superior in the learning of Hungarian inflexion rules.