Hybrid NLP Model for Multilingual Sentiment and Emotion Analysis in Poetry

India is a land of diverse cultures, each with its unique language. As an aspect of literature, poetry applies language to evoke emotions, create imagery, and convey deep meanings, to mean more than what is usually said, hence bridging these cultures. The uniqueness of poetry is that it can show multiple feelings around it within a single stanza. Core emotions like love, joy, sorrow, and hope remain universally expressed despite the linguistic variety. With advancements in Artificial Intelligence (AI), particularly in Natural Language Processing (NLP), machines can now analyze and generate human language. This paper focuses on applying sentiment analysis (SA) and emotion analysis (EA) to poetry, a domain rich in linguistic and emotional complexity. The study examines poems in four languages–English, Hindi, Marathi, and Gujarati–using eight models: Long Short-Term Memory with Attention mechanism (LSTM+ATT), Bidirectional Gated Recurrent Unit (Bi-GRU), Logistic Regression (LR), Support Vector Machine (SVM), Naïve Bayes (NB), ALBERT & GPT-2, RoBERTa & AutoModel, and Neuro-Fuzzy Inference System (NFIS) as the proposed hybrid model. Sentiments are classified as positive, neutral, or negative, while nine emotions are analyzed. This research highlights the models' ability to handle the syntactic, semantic, and cultural intricacies of multilingual poetry and offers a detailed comparative evaluation of their performance.