Abstract:Fractals, Ahead of Print. Currently, Text-to-Speech (TTS) or speech synthesis, the ability of the complex system to generate a human-like sounding voice from the written text, is becoming increasingly popular in speech processing in various complex systems. TTS is the artificial generation of human speech. A classical TTS system translates a language text into a waveform. Several English TTS systems produce human-like, mature, and natural speech synthesizers. On the other hand, other languages, such as Arabic, have just been considered. The present Arabic speech synthesis solution is of low quality and slow, and the naturalness of synthesized speech is lower than that of English synthesizers. Also, they lack crucial primary speech factors, including rhythm, intonation, and stress. Several studies have been proposed to resolve these problems, integrating using concatenative techniques like parametric or unit selection methods. This paper proposes an Applied Linguistics with Artificial Intelligence-Enabled Arabic Text-to-Speech Synthesizer (ALAI-ATTS) model. This ALAI-ATTS technique includes three essential components: data preprocessing through phonetization and diacritization, Extreme Learning Machine (ELM)-based speech synthesis, and Grey Wolf Fractals Optimization (GWO)-based parameter tuning. Initially, the data preprocessing step includes diacritization, where diacritics are restored to unvoweled text to ensure correct pronunciation, followed by phonetization, translating the text into its phonetic representation. Then, the ELM-based speech synthesis model uses the processed dataset for speech generation. ELMs, well known for their excellent generalization performance and fast learning speed, are especially suitable for real-time TTS applications, balancing high-quality speech output and computational efficiency. Lastly, the GWO methodology is employed to tune the parameters of the ELM. The simulation outcomes validate that the ALAI-ATTS technique considerably enhances the intelligibility and naturalness of Arabic synthesized speech compared to existing approaches. The experimental results of the ALAI-ATTS technique portrayed a lesser value of 3.48, 0.15 and 1.37, 0.25 under WER and DER.

Integrating Applied Linguistics With Artificial Intelligence-Enabled Arabic Text-To-Speech Synthesizer

Advancements in Arabic Text-to-Speech Systems: A 22-Year Literature Review

AI-based Arabic Language and Speech Tutor

A Transfer Learning End-to-End ArabicText-To-Speech (TTS) Deep Architecture

ELAICHI: Enhancing Low-resource TTS by Addressing Infrequent and Low-frequency Character Bigrams

Towards Zero-Shot Text-To-Speech for Arabic Dialects

An Expert System for Automatic Reading of A Text Written in Standard Arabic

NatiQ: An End-to-end Text-to-Speech System for Arabic

Arabic Speech Recognition: Advancement and Challenges

Linguistic disparities in cross-language automatic speech recognition transfer from Arabic to Tashlhiyt

ARTIFICIAL INTELLIGENCE IN SIMULTANEOUS INTERPRETING TRAINING: AN EXPERIMENTAL STUDY OF SPEECH-TO-TEXT TECHNOLOGY

A real-time Arabic avatar for deaf–mute community using attention mechanism

Multi-Class Spoken Language Detection Using Artificial Intelligence With Fractal Al-Biruni Earth Radius Optimization Algorithm

FOOCTTS: Generating Arabic Speech with Acoustic Environment for Football Commentator

Effective Deep Learning Models for Automatic Diacritization of Arabic Text

A multilingual training strategy for low resource Text to Speech

ArTST: Arabic Text and Speech Transformer

Hybrid approaches for automatic vowelization of Arabic texts

Exploring Speech Enhancement for Low-resource Speech Synthesis

Accurate synthesis of Dysarthric Speech for ASR data augmentation

Automated Multi-Dialect Speech Recognition Using Stacked Attention-Based Deep Learning With Natural Language Processing Model