Showcase: a Japanese novel, read end to end in Vietnamese

Adding a language is easy to announce and hard to prove. So for Vietnamese we did the same thing we did for English: ran a whole novel through, measured the result, and published the numbers — including an independent quality check.

The book

Kokoro (こころ) by Natsume Sōseki, 1914 — one of the most-read novels in Japan, long out of copyright (the source text comes from Aozora Bunko, the Japanese public-domain library). It's a real stress test: 111 sections across three parts, about 166,000 characters, and a cast the narrator mostly names by role — Sensei, Okusan, Ojōsan, and the friend known only as K.

That cast is what breaks stitched-together translations. Done in disconnected chunks, 先生 lands as "Sensei" in one chapter and "the teacher" in the next. By the second part, the reader is doing detective work.

The run

One EPUB in, Vietnamese selected, no hand-holding: the pipeline read the structure, built a glossary of recurring names, translated all 222 packets with surrounding context in view, and produced a finished Vietnamese EPUB — parts, chapters, and table of contents intact. It ran unattended; the cost was less than a paperback.

What the glossary did

Before translating, Langelic identified the book's recurring people and places and pinned a rendering for each — this run built a 9-entry glossary automatically:

Original	Pinned rendering	Policy
先生	Sensei	transliterate
奥さん	Okusan	transliterate
お嬢さん	Ojosan	transliterate
陛下	Bệ hạ	translate
仏教	Phật giáo	translate

Every packet translated afterwards sees these pins, so section 108 makes the same call as section 3.

The measurements

After the run we queried the finished Vietnamese text directly:

Check	Result
Chapters where Sensei appears	54 of 111
Times "Sensei" appears in the Vietnamese	595
Times "Okusan" appears	391
The character's name swapped for a generic Vietnamese noun	0

The last row needs one honest footnote. Vietnamese leans on relationship words where Japanese leaves the pronoun unsaid, so the narrator often addresses Sensei as thầy — the respectful form for a teacher or elder. That's correct register, not a slip: the character's name is still "Sensei" every time, and where the plain noun for "teacher" appears it is translating a generic Japanese referent (学校の先生, 教師), never the man the book calls 先生.

An independent quality pass

A separate automated reviewer then read the finished EPUB end to end and returned a pass. It confirmed: the recurring names (先生 → Sensei, 奥さん → Okusan, お嬢さん → Ojosan, Ｋ → K) transliterated consistently, chapter numbers and part titles translated (一 → Một, 上 → Phần Thượng), heading anchors well-formed across all 111 chapters, and no Japanese leaking into the Vietnamese text except inside preserved translator's notes. The prose reads as natural, literary Vietnamese.

Read the opening yourself

Here is how the book now begins in Vietnamese — the famous first lines, machine-translated with the glossary in force:

Tôi vẫn luôn gọi người ấy là Sensei. Vì vậy ở đây tôi cũng chỉ viết là Sensei, không tiết lộ tên thật. Điều này không hẳn là sự dè dặt vì ngại miệng đời, mà bởi đối với tôi, gọi như thế là tự nhiên hơn cả.

What we're not claiming

This is a machine translation of a literary classic, honestly labelled. A human literary translator would make a thousand better sentence-level choices. What this run demonstrates is the mechanical promise: structure preserved, names consistent, a real book back on your e-reader — for the vast majority of books that will never get a human Vietnamese edition at all.

Try it on the book you actually want

The novel your feed keeps mentioning. The memoir that never left its home market. Drop in the EPUB or PDF, see the cost estimate up front, and read it tonight.

Translate a book →