It is a rarity that writers at the Digital Orientalist explore projects that might be described as “DH for fun,” but today I’d like to do just that.[1] The Center for Open Data in the Humanities (CODH) launched soan そあん in August this year. Soan is a online service that allows users to convert Japanese text into an image written in cursive Japanese (Kuzushiji くずし字) using data from a dataset of old typefaces (kokatsuji 古活字).
The platform offers two simple options for image creation linked to two different text input boxes that can be accessed through the platform’s webpage. The first is labelled Sāba kyōyūgata sābisu サーバ共有型サービス (Server-Sharing Service) and appears directly on the webpage. Users input their text into the text box, click the Kuzushiji gazō o seisei! くずし字画像を生成！(Generate Cursive Character Image!) button, and are then transported to a new page featuring the generated image, their input text, and the text contained in the image which differs from the input text since in order to generate something that looks like a historical document some kanji and all kana are transformed into hiragana, kana lose their dakuten, and diagraphs appear as full-sized characters. As its name suggests, this option sends the input text and produced image to the server.
The other option is to use the Burauza kanketsugata sābisu ブラウザ完結型サービス (Browser-Based Service). This demo version is access by clicking the hyperlink labelled Soan raiburari no demo そあん（soan）ライブラリのデモ. Unlike the server-sharing service, the browser-based service has explanations in English and Japanese and offers the user the chance to change different parameters to influence image creation. They can, for instance, limit the number ligatures, choose number of characters per line (the default number in the server-sharing service is 15), change line spacing etc. Images created on soan can be used freely.
What is the purpose of this? I hear the reader say. The project’s webpage offers a few potential examples – the first is its possible use in education. By rendering modern Japanese into pre-modern cursive script barriers to learning how to read cursive characters are potentially reduced. The project page argues that working with pre-modern materials can be difficult since it requires a knowledge of pre-modern grammar, vocabularly, and context, and thus the student must decipher a text on several levels. A genreated image, however, can be used to negate these barriers. In other words, through using a platform like this learners are able to learn cursive characters without focusing on textual content. This is showcased in linked projects, such as “Edomi News,” a platform that converts news headlines from Yahoo news into cursive Japanese.
Although I believe that the service may offer the chance to create some interesting materials for a class or two in courses on Japanese palaeography, I don’t think it can replace the use of pre-modern materials. We need to train our students to engage with the content of a text and not only the shape of the characters. After all, it is very unlikely that they will engage with cursive texts written in modern Japanese in the real world or within their careers as students and scholars. Furthermore, the clues we gain from content are, in fact, an essential part of the process of deciphering cursive characters and historical Japanese texts. Though it is likely useful as a tool to create some novel additions to a class, soan is perhaps not as revolutionary as an educational tool as suggested. Nevertheless, where learners do not have access to pre-modern materials using soan may help to provide greater access to cursive texts. Soan’s webpage also suggests that since images from the service can be shared, it may encourage communication in cursive characters.
Soan is an interesting use of data and a fun service to try out. It will undoubtedly allow educators to dabble in creating some extra materials for their classes, but is unlikely to revolutionize the study of Japanese palaeography and methods of learning cursive characters. Check it out here for yourself!
[1] The term “DH for fun” is being used to describe my own interpretation of soan. I do not presume to have divined the goals of the developers.
The Digital Orientalist 
One thought on “Generating Kuzushiji Images with Soan”