Maintaining Translation Consistency
Mercari, Inc. operates Japan’s largest marketplace app, Mercari, along with the mobile payment service Merpay. Over the last few years, Mercari has grown from a mainly Japanese company to having an extremely diverse Tokyo office, with employees from all over the world. Many employees who have joined the company from overseas have limited Japanese language skills, so Mercari has an in-house translation and interpretation team dedicated to providing language support to bridge the gap between Japanese speakers and English speakers.
Bridging the language gap in a multilingual team
Previously, the demand for translation was small, and the translation team only needed a few members to handle everything. However, as the number of non-Japanese speakers rapidly increased, so did the number of translation requests, and the translation team began expanding as well. It wasn’t long before they identified a growing problem: translation consistency.
As the translation team works on a request basis, rather than a project basis, related documents are sometimes assigned to different translators. Even with glossaries, translators tend to translate the same terms differently; you may think to reference a glossary for proper nouns like product names, but not so much for everyday words like verbs or adjectives, even though inconsistent translations of these can cause confusion.
On top of that, the translation team found themselves doing the same translation multiple times—a document would be updated and translation requested again, but unless someone remembered that it had been translated once before, the document would be translated again from scratch, resulting in unnecessary work and wasted time.
Translation memory and term bases to speed up the translation process
Mercari introduced Phrase TMS in 2019. After building a translation memory (TM) from past translations, Phrase TMS almost immediately showed positive results; weekly updates, for example, are automatically filled in with all of the segments that were unchanged from the previous week. If the translation team receives a translation request that may have been done before, all they have to do is upload it to Phrase TMS, and see whether the segments are in the TM already.
In terms of consistency, the term base suggestions automatically showing up for each segment removes the work of having to reference a glossary, and puts the information right at the translator’s fingertips. Even if something doesn’t show up in the term base, the translator can easily search the TM for all instances of a word and see how it’s been translated in the past.
Phrase has transformed our localization process and we started to see positive results from day one. We are able to measure and analyze the scope of each translation project enabling accurate and instant insight for all stakeholders.
Measuring and analyzing translations
Phrase TMS also enables Mercari to measure and analyze their translations in a way they could not before. Many of their translations are done on Google Docs, Sheets, and Slides, but Sheets and Slides do not offer built-in solutions to measure character/word count. By uploading files to Phrase TMS through the Google Drive connector, Mercari’s translators can see not only an accurate character count of all documents, but a breakdown of how many segments/characters are already in the TM. The team is currently using this to improve their workflow and more accurately estimate how long translations will take based on match rates.