Hacker News new | ask | show | jobs
by planbattack 1164 days ago
Japanese to English translator here. The general rule of thumb (that is often used for billing estimates) is that N Japanese characters = N/2 English words.

So if you have a Japanese source text that is 2,000 characters, the English translation will be around 1,000 words.

I tested a translation (one sentence) from a previous job:

Japanese: 94 characters, 128 tokens

English: 39 words (232 characters), 47 tokens

Seems quite unbalanced given that the amount of "information" in the two is equivalent.

1 comments

Oof... that's a rough job to have in the world of ChatGPT...