Code and data for our Figlang@EMNLP2022 paper: The Secret of Metaphor on Expressing Stronger Emotion

Explaining data files

specificity.tsv annotates which is more specific? metaphor or literal.

• The specificity column indicates whether the metaphor is more specific. 1 means metaphor is more specific, 0 means literal is more specific, and 2 means same specific.
• The common_hyper column indicates the common hypernym of the metaphoric and literal term. For example, (Synset('blast.v.07'), 1, 0, Synset('blaze_away.v.02'), Synset('blast.v.07')), the first item of the tuple indicates their common hypernym, the 1 and 0 means it takes the metaphoric term 1 step to reach the common hypernym, and it take 0 steps to reach the common hypernym. the last two items indicate the synset of metaphor and literal terms.

similar_specific_terms.tsv annotates which is more emotional? metaphor or more specific literal.

• The more_emotional column indicates whether the chosen more specific literal term is more emotional. 2 means the metaphor is more emotional; 1 means the more specific literal is more emotioanl; and 0 means they share the same level specificity.
• The substitute column is the manual chosen more specific literal term.
• The similar_specific_terms are all potential terms from worknet sharing the same level specificity with the metaphoric one.

more_specific_terms.tsv annotates which is more emotional? literal or more specific literal.

• The substitute column is the manual chosen more specific literal term.
• The emotional column indicates whether the chosen more specific literal term is more emotional than the original literal one. 2 means the original literal is more emotional; 1 means the more specific literal is more emotioanl; and 0 means they share the same level specificity.

Expalining code files

wordnet_level.py locate terms in the wordnet hierarchy.

specificity_emotion.py produce terms in the same/lower level specificity in the wordnet hierarchy.