Reproduce “Annotation and modeling of emotions in a textual corpus: an evaluative approach” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jonas Noblet

A study introduces a new French corpus of user opinions on innovative concepts, annotated with an evaluative emotion model (EMINOSA). It demonstrates that fine-tuned Transformer-based language models can effectively reproduce the gradient-like nature of human evaluative emotion annotations, even with significant inter-annotator variability.

Key insights

  • Inter-annotator disagreement in emotion annotation can reflect a stable "gradient of intensity" in emotional expression, where variability is a signal rather than just noise.
  • The evaluative emotion approach, focusing on appraisal dimensions, offers a nuanced framework that is well-suited for analyzing reactions to specific stimuli like innovative concepts.
  • Advanced language models can be trained to capture and reproduce the subjective, gradient-like nature of human emotion annotations, even with high inter-annotator variability.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.