Masked Metaphor Modeling To Transfer Literal to Metaphorical Text

This study presents a new approach to metaphorical paraphrase generation by masking literal tokens of literal sentences and unmasking them with metaphorical language models. Unlike similar studies, the proposed algorithm is not limited to the replacement of verbs, but also of nouns and adjectives. Despite the fact that the transfer rate for the former is the highest (56%), the transfer of the latter is feasible (24% and 31%). Human evaluation showed that our system-generated metaphors are considered more creative and metaphorical than human-generated ones. Additionally, when using our transferred metaphors for data augmentation we show that state of the art metaphorical sentence classification improves by 3% in F1.