Patrick Thomasius


2023

pdf bib
Sam Miller at SemEval-2023 Task 5: Classification and Type-specific Spoiler Extraction Using XLNET and Other Transformer Models
Pia Störmer | Tobias Esser | Patrick Thomasius
Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023)

This paper proposes an approach to classify andan approach to generate spoilers for clickbaitarticles and posts. For the spoiler classification,XLNET was trained to fine-tune a model. Withan accuracy of 0.66, 2 out of 3 spoilers arepredicted accurately. The spoiler generationapproach involves preprocessing the clickbaittext and post-processing the output to fit thespoiler type. The approach is evaluated on atest dataset of 1000 posts, with the best resultfor spoiler generation achieved by fine-tuninga RoBERTa Large model with a small learningrate and sample size, reaching a BLEU scoreof 0.311. The paper provides an overview ofthe models and techniques used and discussesthe experimental setup.