System for identifying and correcting invalid words in the devanagari script for text to speech engine

Abstract

The Text To Speech (TTS) system takes text as an input and generates speech as an output. If input text is incorrect then overall quality of speech output may degrade. The main aim of the proposed system is to provide correct input text to the TTS. The system takes Unicode word as an input, identifies invalid word and corrects it by inserting, deleting or updating characters of the word. In this system, the State Machine is used to identify and correct invalid word in the Devanagari script which in turn is based on rules. Rules are developed for converting character to input symbol. Actions and States are identified for State Machine. Finally, the state transition table is developed for validation and correction of word. Using this system, incorrect words of the Devanagari script can be corrected to valid words (word contains all the valid Devanagari syllables) based on Devanagari script grammar. Since, all Devanagari characters are not present in Hindi language; this system will correct these nonHindi characters to Hindi.