| Resumen |
The present work aims to present different methods for detecting humor in One-liners, in general humor is everything that causes us gracefulness or that in an ingenious way presents us with a humorous punchline. In this way, it is common to find humor that its construction tends to be related to the topic being discussed, its premises / punchlines. To classify humor in texts there are several aspects: A powerful embedding that correctly represents the meaning we want to obtain, a model robust enough that it is easy to learn with our data and finally a combination of both an embedding / robust model capable of successfully carrying out the expectations of the given task. As a primary approach, pre-trained embeddings were used in a basic CNN in contrast to the paradigm of Tranformers. Obtaining good results in both areas for both embedding and pre-trained transformer models, with a qualification above 99 of the F1-Score. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG. |
