Is a machine WORD always the same or does it depend on the machine architecture? And is the meaning of the word WORD context sensitive or generally applicable?
Yes.
Ok, let me be a bit clearer. Assuming we are talking about words of memory, there are two broad definitions.
First, a word is often the natural size of a single item that can be accessed atomically in the hardware. That is very much a platform dependent size, but is usually 16, 32, or 64 bits, but other sizes have been found in the wild.
Second, it is often used to specifically mean a 16-bit value. In that context, you will see DWORD used to mean a 32-bit value. This usage is common on x86 platforms, especially Windows, but was used on DEC PDP-11 and VAX, and Motorola 68000 descendants as well.
Telling which is the intended usage depends on context...