Reproduce “Representing Affect Information in Word Embeddings” 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
Yuhan Zhang, Wenqi Chen, Ruihan Zhang, Xiajie Zhang
The authors compare four word representations to human ratings of valence, arousal, and dominance, using classifiers and similarity probes. They examine static GloVe, base BERT, GoEmotion-fine-tuned BERT, and a contextualized BERT vector built from movie-review occurrences. Contextualized BERT reaches 0.90 dominance accuracy versus base BERT’s 0.82, but the probes do not establish downstream application gains or causal effects of context.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.