University of Twente Student Theses
Dissecting humor to find LLMs funny bone- Large language models recognition of humor based on syntactic ambiguity
Konter, J.T. (2025) Dissecting humor to find LLMs funny bone- Large language models recognition of humor based on syntactic ambiguity.
|
PDF
446kB |
| Abstract: | This study investigates the ability of large language models (LLMs) to recognize humor based on syntactic ambiguity, a phenomenon where a sentence can be interpreted in multiple ways due to its structural complexity. Through a comprehensive literature review, we explore the relationship between syntactic ambiguity and humor, focusing on how LLMs comprehend and process these linguistic nuances. We find that humor arising from syntactic ambiguity often relies on the subversion of expected interpretations, as explained by Raskin’s Semantic Script Theory of Humor (SSTH) and its extension, the General Theory of Verbal Humor (GTVH). Regarding LLMs, our analysis reveals that while these models exhibit partial understanding of syntactic structures and ambiguity, their performance is inconsistent and heavily dependent on training data. Additionally, LLMs' syntactic representations are not robust. Few-shot learning improves their ability to recognize ambiguity, but their processing mechanisms differ significantly from human cognition. Based on these findings, we propose that enhancing LLMs' ability to detect syntactic ambiguous humor could involve incorporating more examples of such humor into training data and leveraging semantic cues. Future research could explore the impact of humor theories on LLMs' performance and further investigate the interplay between syntactic and semantic processing in these models. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 17 linguistics and theory of literature, 50 technical science in general |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/105113 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page