Historically, messages in languages that use alphabets have been encrypted manually according to some kind of algorithm (e.g. mono- and poly-alphabetic ciphers). But how were messages encrypted in a semagram-based language like Chinese?
There was a telegraphic code that mapped the comparatively frequently used 10000 words (that's very much more than for common daily use, e.g. newspapers) of Chinese to [0,9999]. There was once also a patented mechanical device that effected a poly-alphabetical substitution of [0,9] with a set of disks containing certain scrambled alphabet of [0,9], which the user selected 4 of them (aggreed upon with the partner) to encrypt the 4 digits from the telegraphic code. Since frequency analysis of the digits barely helps, the result from that device isn't bad IMHO even today (anyway assuming one has a somewhat larger set of disks than at that time.)
The Chinese used digits. So there is an encoding into digit groups, like a code book. This code book could be standard (like the Chinese used for telegraphy and radio as well, which used a publicly available book), or for extra secrecy could be kept a secret. The cipher could then be made to work on digit groups (and there standard techniques like addition mod 10 with a key stream, permuations, etc could be used).
The Japanese also used digit based codes a lot, or they used (in their WW2 era machines) the transliteration (there are only a finite number of syllables) into a Latin alphabet, and then used rotor machine techniques (as for the enigma) etc.
But the digit route seems the more common.