AI for Book Translation
 |
AAI for book translation works. Perhaps not as well as it works for automated audiobook creation. But it’s getting very close, very fast. Non-literary fiction could be first. Literary fiction may follow. Nonfiction poses a different set of challenges. |
I hosted a webinar on AI for book translation, sponsored by BISG, in June, 2024. The video is online on YouTube. Jane Friedman also described the program in her Hot Sheet newsletter.
The subject is complex and nuanced. One thing I find fascinating is how long people have been trying to automate translation. It’s a reminder that books, which fill our universe, are such a small proportion of written communication, even more so in this online age.
Warren Weaver, credited as the father of machine translation (MT), noted to a colleague, “When I look at an article in Russian, I say: ‘This is really written in English, but it has been coded in some strange symbols. I will now proceed to decode.’” For a machine, language is just code. It’s not culture and feeling and the grandeur of written language. It’s a task with numbers and code.
Clearly the fiction/nonfiction divide will loom large in AI translation. Chat AI is strong on style, but it falls short on facts. It’s very early days, but I’m inclined to think that Chat AI will shine with fiction, but come up short with nonfiction. (Nonetheless, I’m proceeding to machine-translate this fact-filled book!)
Literary fiction is very much the elephant in the room. It is precious and revered and rightly so. Translators can spend hours arguing about a single word or phrase. Chat AI must tread cautiously in those waters.
But this is fertile territory. As far as I can determine (from scant data), there were only 9,500 trade book translations in 2023. Even if I’m off by a large factor, it’s clear that few books are being translated from foreign languages into English.
Similarly, I found a statistic indicating that in 2023 there were only 7,230 translations from English into Spanish (in Spanish book markets). That seems ludicrously small.
There’s a vast opportunity here.
Most of the use of AI for book translation will be for books where translation was never considered economically feasible. There is bound to be a job impact on translators of “mid-market” books; the job growth will be managing projects and in QA. Will that offset the job loss? Unlikely.
As with most aspects of AI, there are challenging issues to be addressed, and no easy answers.
As I note in the software section, Leanpub and DeepL are the two companies offering AI-assisted book translation services to authors and publishers.