Reproduce “FrameBERT: Conceptual Metaphor Detection with Frame Embedding Learning” 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
Yucheng Li, Shun Wang, Chenghua Lin, Frank Guerin, Loïc Barrault
FrameBERT, a model developed by researchers from the University of Surrey, University of Sheffield, and Meta AI, explicitly incorporates FrameNet embeddings into a RoBERTa-based architecture for conceptual metaphor detection. The model achieves state-of-the-art or comparable F1-scores on benchmark datasets and offers enhanced interpretability by revealing the conceptual frames informing its predictions.
Key insights
- Explicitly integrating structured conceptual knowledge from FrameNet significantly enhances metaphor detection capabilities beyond what 'shallow' contextual embeddings alone can achieve.
- The learned conceptual embeddings are critical for FrameBERT's performance, as evidenced by a substantial F1-score drop of 13% when they are randomized during evaluation.
- The model provides increased interpretability by using distinct literal and contextual semantic frames to explain its classifications, demonstrating the underlying conceptual reasoning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.