Using NotebookLM can make the rough drafting process for literary translation easier.
Reference: Utilizing AI in Novel Writing—NotebookLM.
Literary translation is an extremely difficult task, considering the time and cost involved. There are many excellent works of Japanese literature, past and present, but it is a pity that they are unknown to the world simply because they have not been translated. If NotebookLM can make this task even slightly easier, it holds great significance. It might even become possible to translate and introduce works from Aozora Bunko (or similar collections) into English or other languages.
Translation can be executed using either NotebookLM’s chat or report functions, but for long texts, the report function is likely better. Even with a simple prompt like “Translate into Japanese,” it can provide translations accompanied by literary analysis (if analysis is unnecessary, you can specify “without commentary”). Generally, sentences of a certain length where the context is clearer are translated better than fragmented sentences. However, generating the report itself may fail when dealing with extremely long texts, such as novels.
Now, what about the results? Let’s take Charles Dickens’ A Christmas Carol as an example. Since more than 100 years have passed since the death of this great author, the copyright protection period has expired, making it part of the public domain and freely available to anyone. This is what the story is about.
Below is part of the opening. The original text is from The Project Gutenberg.
“Don’t be cross, uncle!” said the nephew.
“What else can I be,” returned the uncle, “when I live in such a world of fools as this? Merry Christmas! Out upon merry Christmas! What’s Christmas time to you but a time for paying bills without money; a time for finding yourself a year older, but not an hour richer; a time for balancing your books and having every item in ’em through a round dozen of months presented dead against you? If I could work my will,” said Scrooge indignantly, “every idiot who goes about with ‘Merry Christmas’ on his lips, should be boiled with his own pudding, and buried with a stake of holly through his heart. He should!”
“Uncle!” pleaded the nephew.
“Nephew!” returned the uncle sternly, “keep Christmas in your own way, and let me keep it in mine.”
“Keep it!” repeated Scrooge’s nephew. “But you don’t keep it.”
“Let me leave it alone, then,” said Scrooge. “Much good may it do you! Much good it has ever done you!”
If you make NotebookLM translate it, the result is as follows.
「怒らないでくださいよ、おじさん！」と甥は言った。
「他にどうなれるというのだ」と叔父は返した。「こんな馬鹿ばかりの世界に住んでいて。メリー・クリスマスだと！ メリー・クリスマスなんてくそくらえだ！ お前にとってクリスマスなんて、金もないのに請求書の支払いをする時、一歳年を取るだけで一時間も豊かにならないと気づく時、帳簿を締め、その中のあらゆる項目が丸一年分、お前に不利な証拠として突きつけられる時、それ以外に何だというのだ？ もし私の思い通りにできるなら」とスクルージは憤慨して言った。「『メリー・クリスマス』と口にして歩き回る馬鹿は皆、自分のプディングで茹でられ、心臓にヒイラギの杭を打ち込まれて埋められるべきだ。そうすべきだとも！」
「おじさん！」と甥は嘆願した。
「甥よ！」と叔父は厳しく返した。「クリスマスはお前のやり方で祝い、私には私のやり方で祝わせろ」
「祝うですって！」とスクルージの甥は繰り返した。「でも、おじさんは祝っていないじゃないですか」
「では、放っておいてくれ」とスクルージは言った。「それがお前に大いに役立つといいな！ これまで大いに役立ってきたのだろう！」
It’s done quite well. The tone is consistent, and it largely handles the omission of subjects. Translating “keep Christmas” as 「クリスマスを祝う」 (to celebrate Christmas) is also a good approach.
However, even setting aside the minor points, the following issues are apparent when viewing other sections:
1. The translation lacks polish and needs improvement.
2. There are unnecessary repetitions.
3. Confusion in self-reference and references to others (rarely, the way others are addressed may change).
4. The translation of proper nouns can sometimes be inappropriate.
5. The translation may not be natural.
Several things can be done to further improve the translation quality:
• Create a glossary: Check and revise the results. Example prompt: “Create a bilingual glossary”.
• Create a translation table (alignment): Divide by sentence and compare the source and target languages in a table format for review. This was formerly known as “translation memory”.
After creating these, add them to the source documents in Markdown format and execute the translation again. The advantage of NotebookLM is that you can provide feedback to the system to achieve better results. For long works, dividing them into shorter sections and checking the quality is also effective.
Ultimately, if one seeks better results, the true value of the tool may only be realized when used by a translator who has the ability to properly review the output, rather than relying on it completely. Performance is expected to improve in the future, however.