When choosing how to read a compound word, there’s the decision of what reading each character gets (usually on, sometimes kun, and sometimes which on or kun to use). After that, there’s the decision of whether rendaku applies. That is, there’s 1 bit of information: does it rendaku, yes/no?
I have read Tofugu’s article, studied Lyman’s Law. I am familiar with many always-rendaku and never-rendaku words, searched these community pages and read through many discussions. These give nice heuristics, but at the end of the day it comes down to exposure or rote memorization.
I’m interested if others have attempted to deploy any mnemonic-like techniques on a case-by-case basis for rendaku. That is, for words where it is not obvious from above listed heuristics and prior exposure.
For example, if using a memory palace-like technique, one could think of the words as occupying some specific spacial relation to oneself (e.g. rendaku exceptions are always behind me, or to my left). Or, one could use a temporal relation (e.g. the exceptions are associated with feudal Japan in my mind, or the distant future). Or, one could even ascribe a consistent categorization such as a color or image or association with a person or object.
Has anyone done anything beyond exposure, rote memorization, and heuristics?