Supriyono, Aji Prasetya Wibawa, Suyono, Fachrul Kurniawan
Summarizing stand-up comedy scripts presents a unique problem because of the complex interplay of language, timing, and cultural references that define humor. Traditional summarization techniques sometimes fail to preserve the punchlines, irony, and narrative coherence essential to comedic storytelling. This study introduces a hybrid methodology that integrates linguistic cohesion and coherence methodologies with a deep learning framework employing Long Short-Term Memory (LSTM) networks and Word2Vec embeddings. Our objective is to develop a summarization model that captures both the structural coherence and humorous essence of stand-up performances in a more succinct format. We employed a bilingual dataset of English and Indonesian comic scripts for text preprocessing, embedding, and sequence modelling during system training. The assessment employed metrics such as accuracy, precision, recall, and F1-score, together with a qualitative analysis of the generated summaries. The results demonstrate that the proposed methodology significantly improves humor retention and logical coherence compared to conventional methods. The accuracy increased from 0.65 to 0.83 after training, and validation measures confirmed strong generalization. This approach offers considerable promise for content regulation, customized recommendations, and AI-augmented entertainment solutions. This research improves natural language comprehension in informal, culturally diverse genres by preserving the distinctive structure and intent of comedic language. © 2026, Politeknik Negeri Padang. All rights reserved.
Department of Electrical Engineering and Informatics, Faculty of Engineering, Universitas Negeri Malang, Malang, Indonesia; Department of Indonesian Literature, Faculty of Letters, Universitas Negeri Malang, Malang, Indonesia; Faculty of Science and Technology, Universitas Islam Negeri Maulana Malik Ibrahim Malang, Indonesia
