Papers with task-based evaluation
MERMAID: Metaphor Generation with Symbolism and Discriminative Decoding (2021.naacl-main)
Copied to clipboard
| Challenge: | a new method for generating metaphors is proposed to generate literal sentences . human evaluations show that our best model generates metaphors better than three well-crafted baselines 66% of the time on average. |
| Approach: | They propose a method to automatically construct a parallel corpus by transforming literal sentences to metaphorical ones using commonsense inference and masked language modeling. |
| Outcome: | The proposed method generates metaphors better than baselines 66% of the time on average. |