Automatic Transcription Of English Connected Speech Phenomena

Abstract

Many phonetic phenomena that occur in connected speech are classified as phonetic periphery where anything can happen. A well-known convenient way to fix any phonetic phenomenon using certain symbols is transcription. The current paper aims at showing the model of predicting allophones by coordinating a number of factors that determine the choice of a particular allophone and visualizing the result changing certain letters into corresponding IPA symbols. Free Pascal compiler and Geany editor are used for programming purposes. The model is created for American English. It is tested for tap and glottal burst, the latter being one of the three glottalization patterns. The difference of the combination of factors for purely linguistic analysis and for computer programming is explained. We demonstrate (i) the framework for integrating separate blocks each dealing with one phenomenon (ii) a block for tapping which is almost finalized and a part of a block on glottalization, particularly patterns for glottal burst.