In Pint you can define custom units in a unit registry. So you make your complex unit out of predefined units and the underlying unit conversion is built-in. Pint also works with pandas/numpy so you you get vectorized operations too for performance.
I understand if you are getting a human to translate each sentence you can just point them to the right phrases but my understanding is that you're translating text in bulk via machine translation. For scales in which the conversion is linear and the starting point is zero, this task is relatively trivial. For scales like temp where the zero points are different, you need to treat deltas different from absolute values. For scales where one is the inverse of the other, there's a lot of very tricky edge cases that make machine translation hard.