Detecting Implicitly Quoted Waka within 'The Tale of Genji' through Vector Search

Abstract

This study investigates the rhetorical technique known as "Hikiuta" (implicit Waka) in Japanese classical literature, which involves the implicit use of Waka poems from preceding works. The aim was to digitally and comprehensively identify these references. Focusing on "The Tale of Genji," we created a database of sentence vectors for the entire text. We then matched these vectors against those of all poems in the "Kokinshu" through nearest neighbor searches, selecting pairs with closely matching vector values. By further filtering these results based on the high degree of match in character N-grams, we were able to detect "Hikiuta" with remarkable precision. This methodological approach not only enhances our understanding of intertextual references in Japanese classical literature but also showcases the potential of digital tools in literary studies.