@inproceedings{parthiban-r-2026-humor,
    title = "Humor Generation {--} Text-based Humor Generation ({E}nglish)",
    author = "Parthiban, Hemeshkumar  and
      R, Priyadharsini",
    editor = "Kochmar, Ekaterina  and
      Ghosh, Debanjan  and
      North, Kai  and
      Komachi, Mamoru",
    booktitle = "Proceedings of the 20th {I}nternational {W}orkshop on {S}emantic {E}valuation (2026)",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.semeval-1.104/",
    pages = "735--740",
    ISBN = "979-8-89176-414-9",
    abstract = "Our system for SemEval-2026 Task 1 Subtask A addresses constrained text-based humor generation in English. The approach relies on structured prompt engineering using a GPT-4{--}class large language model in a zero-shot setting without task-specific fine-tuning. Each input, consisting of either mandatory word pairs or a news headline, is embedded into a fixed instruction template enforcing strict stylistic and structural constraints.The system ensures single-sentence outputs between 8{--}12 words, adopts a dry and deadpan tone, and incorporates subtle expectation shifts while avoiding exaggerated punchlines or unsafe content. Deterministic decoding guarantees replicability, and an automatic validation step enforces compliance with official submission requirements.Experimental results show that structured prompting significantly improves stylistic alignment compared to unconstrained generation. The system demonstrates that controlled humor generation can be achieved through constraint-based prompt design without additional training."
}