JCT at SemEval-2026 Task 1: Let the Best Joke Win - A Generate - and-Rank Approach to Constrained Humor
Year: 2026
Type: conference-paper
Abstract: We present a humor generation system for SemEval-2026 Task 1, Subtask A (Castro et al., 2026) that produces short jokes under lexical or headline-based constraints.For each input, our system generates multiple candidate jokes using a large language model across diverse humor styles and prompting strategies, including zero-shot, few-shot, and structured prompting.Constraint satisfaction is explicitly enforced, either by requiring exact lexical inclusion or by approximating semantic relevance to a headline using sentence-embedding similarity.All valid candidates are ranked using a weighted humor score that combines semantic incongruity, emotion-based humor potential, irony likelihood, linguistic fluency, and novelty with respect to a large external jokes corpus, and the single highest-scoring joke is selected for each constraint.This approach follows a bestcandidate selection paradigm, leveraging automated humor proxies to improve joke quality without task-specific fine-tuning.
Institution: Jerusalem College of Technology
Language: English
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7166798597