Modeling Word Emotion in Historical Language: Quantity Beats Supposed Stability in Seed Word Selection

  • 2019-04-08 11:32:41
  • Johannes Hellrich, Sven Buechel, Udo Hahn
  • 0


To understand historical texts, we must be aware that language -- includingthe emotional connotation attached to words -- changes over time. In thispaper, we aim at estimating the emotion which is associated with a given wordin former language stages of English and German. Emotion is representedfollowing the popular Valence-Arousal-Dominance (VAD) annotation scheme. Whilebeing more expressive than polarity alone, existing word emotion inductionmethods are typically not suited for addressing it. To overcome thislimitation, we present adaptations of two popular algorithms to VAD. To measuretheir effectiveness in diachronic settings, we present the first gold standardfor historical word emotions, which was created by scholars with proficiency inthe respective language stages and covers both English and German. In contrastto claims in previous work, our findings indicate that hand-selecting smallsets of seed words with supposedly stable emotional meaning is actually harmfulrather than helpful.


Introduction (beta)



Conclusion (beta)