All questions answered.

What?

marinaMoji is a Japanese input method designed for scholars working with premodern and historical texts.

It extends the capabilities of modern Japanese IMEs with tools for historical kana orthography, kyūjitai and shinjitai conversion, kaeriten reading marks, odoriji repetition marks, and other features commonly required in the study of classical Japanese, kanbun, and East Asian textual traditions.

marinaMoji is based on Mozc and is developed as an open-source research software project.

Why?

Existing Japanese input methods are optimized for contemporary Japanese writing. Scholars of premodern East Asia often rely on cumbersome workarounds, custom dictionaries, manual character conversion, and specialized software.

marinaMoji seeks to provide a unified environment for scholarly text input that is free, open source, and compatible with all modern operating systems and applications.

Who?

The project is developed by D.P. Morgan (CNRS) and M. Pandolfino (EPHE) at the CRCAO, with institutional support from the partners listed in the site footer.

Source code, issue tracking, and development discussions are available on GitHub.

How?

marinaMoji is a fork of Mozc, the open-source Japanese input method derived from Google's Japanese Input. It also builds upon the wider open-source ecosystem, including OpenCC, IBus, Qt, libsodium (encrypted sync), miniz (sync bundle format), and many other projects. We thank the maintainers of these projects, as well as the colleagues and early adopters who tested preliminary releases.

Hosting provided by Huma-Num.

Content management by Grav.

Data, security, and privacy

An input method has access to every keystroke that passes through it, including potentially sensitive information such as names, passwords, and bank card numbers. For this reason, input methods are subject to heightened scrutiny by operating systems such as macOS, Windows, Android, and iOS.

Several aspects of Mozc led us, as security- and privacy-conscious developers, to adopt it as the basis of marinaMoji and to use it in our own daily work.

First, Mozc is a mature and actively maintained project derived from Google's Japanese input technology. It has been used and scrutinized by a large community of developers and users over many years.

Second, both Mozc and marinaMoji are open-source software. Their source code is publicly available and can be inspected, audited, modified, and redistributed. While open source does not automatically guarantee security and privacy, it does allow independent verification that the software behaves as documented.

Mozc and marinaMoji are designed to operate entirely offline for text conversion and dictionary lookup. No internet connection is required for normal operation. The macOS version may periodically check for software updates, but text entered through the input method is never transmitted as part of that process.

Mozc and marinaMoji store user preferences, custom dictionary entries, and candidate choice history locally in order to improve user experience. The stock Mozc has a 'send usage statistics' option, but this applied only to official Google/Android builds, and we have further disabled it. marinaMoji does not send usage statistics to Google or to the marinaMoji developers; all user data thus remains on the user's device. We have built a tool to synchronise user dictionaries and candidate choice history between devices (below), but synchronisation is opt-in, encrypted, and based on third-party services rather than infrastructure operated by Google or by the marinaMoji developers.

In the IME menu, toggle Privacy mode to enable incognito-style behaviour: no new history learning while it is on. Use this on shared machines or sensitive drafts. Full settings remain under Properties → Privacy (incognito and history-learning options). Optional encrypted cross-device sync is documented in the Guide.

Dictionary and lexical data

marinaMoji is beginning to ship supplementary dictionary modules alongside the stock Mozc lexicon — for example Japanese era names (gengō) and imperial reign names — to support scholarly typing of proper nouns that general-purpose IMEs often omit.

Where these additions draw on structured data from Wikidata, that source material is dedicated to the public domain under CC0 1.0 (Creative Commons Zero). marinaMoji adapts such entries into its Mozc dictionary format (readings, surfaces, part-of-speech tags); curated tables may also be hand-edited before release.

Other planned lexical sources (JMdict, historical UniDic, project-specific lists) carry their own licences; see the dictionary repository for developer notices.