Emotionally Consistent English to Tamil NMT for Alliterative Text Through MOGA and Pareto Optimization

With the increase in neural networks, automated machine translations are now considered a necessity. Despite their growth, machine-based translations have multiple hurdles when dealing with artistic, imaginative, or literary expressions. The reason is the heavy dependency of the message on patterns, such as alliterations and emotional depths. As the original meaning relies on aesthetics and expressions, the current translation techniques fail to capture the essence and focus only on literal and grammatical precision. The proposed technique gives importance to phonetic patterns, grammar, literacy meanings, and maintaining emotional integrity as well as semantic coherence. A novel method to enhance English-to-Tamil machine translation involves the utilization of emotion-specific embedding and multi-objective genetic algorithms (MOGA) with Pareto-based optimization. The proposed NMT model outperforms the baseline NMT model with transformers achieving 85% semantic accuracy, 80% phonetic fidelity, and 90% emotional consistency.