Abstract
This article focuses on computational analysis of Tang dynasty “poems on things” (yongwu shi 詠物詩) and some of the most common objects described in them. Modern technology offers many possibilities for new approaches to the study of poetic language, and this article discusses some of the tools that can aid in semantic analysis of separate words or poems. These include so-called word embeddings, vector representations of word semantics, and “fingerprints” that are calculated on the basis of word embeddings to represent semantics of whole texts. Applied to classical Chinese poetry, they can show some of the paradigmatic groups of images and their distribution between concepts of happiness and sadness, loneliness and companionship. Finally, topical grouping of poems on things is discussed and explored with the help of fingerprints to look for formal principles behind the grouping of the texts.
