Segmentation of Punjabi Text into Prosodic Unit

Abstract

Naturalness of synthesized speech is defined in terms of accurate pronunciation of words, whether the stress is given to the right word and are the nasal sound words are spoken properly. It is the task of text analysis phase of speech synthesizer to generate linguistic representations of written text by considering nasality, stress and pronunciation of words. As there is difference in written and verbalization form of text. So, text analyser performs text normalization i.e. some preprocessing on written text to transform text into its verbalization form. Some pre-processing tasks performed are date expansion, numeral expansion, alphanumeric expansion, abbreviation expansion, and phonological translation. The vowel-consonant sequence of normalized text is generated that is used to guide the process of segmenting the text into prosodic units (graphemes) and generation the linguistic representation of text.