Learning Folktale Plot Representations

Anonymous

Abstract: Folktales possess both historical and literary significance as they offer a glimpse into the culture and traditions of the communities that created and passed them down through the generations. Folklore scholars have been meticulously analyzing and classifying folktales based on their type and motifs, however, the automatic identification and discovery of folktale types remains an area of ongoing study. We propose a computational approach for identifying folktale types by utilizing an online library of folklore texts and a neural embedding model. Our method semantically encodes the texts, resulting in tale representations that capture the similarities between tale plots. We validate this by visualizing the representations using t-SNE and applying K-means clustering, along with human evaluation.

Paper Type: long

Research Area: NLP Applications