Many cultural memory institutions provide online access to their collections. Ukiyo-e, Japanese traditional woodblock print, is known as a popular art of the Edo period (1603–1868). There are many copies printed from the same ukiyo-e woodblocks, and they have been digitized and exhibited on the Internet by many museums around the world with descriptive metadata in various languages. This means that identical ukiyo-e prints can exist in different digital collections, along with metadata in different languages. This proposal aims at finding these identical ukiyo-e prints by comparing their metadata across languages.
We measure the metadata similarities across languages by utilizing bilingual word embeddings, that allow measuring word similarities across languages. Our method can be used to link or integrate digital collections in different languages, which could help people to explore ukiyo-e prints regardless of their metadata’s languages.