There have been a great deal of discussions throughout the past few years over whether or not Text-to-Speech (TTS) machines can synthesize natural synthetic speech from texts. Aiming to address this need, in this paper we have considered some significant viewpoints of Spoken Language Processing (SLP) on the phonetic transcription of each word for the preprocessing of text-to-speech synthesis. On the other hand, lack of researches on automatic language detection for text transcription in different languages by considering the phonology of that language, motivated us to make a text language identifier system for commonly-used contexts. It therefore encouraged us to conduct a novel research into semi-English texts that people from different nations often use in their conversational transcripts such as email or chat via the Internet or cell phone text messaging services. In this research, we have investigated the language of text sequences by employing phonotactics rules, chiefly on Finglish (a portmanteau term combining Farsi and English) which is an alternative writing format for the Farsi Language, which has its own orthographic system, by means of English letters. As this is the first paper regarding Finglish texts, it can be tremendously exciting to enhance the text-to-speech synthesis systems created by advanced Digital Signal Processing (DSP) algorithms to specifying the language of each sentence in the first place. Finally, we have also proposed highly recommended writing rules for Finglish for it to be easily understood, translated to English and converted to natural speech.

Keywords