The first time a word is encountered, it needs a quick translation, or ,better, a picture or a gestural explanation, especially if it is somewhat abstract.
After that, the brain must be left alone to build its own semantic web for that word. This is a chaotic process, with constant revisions taking place, with meanings constantly shifting and adjusting. Authors can use words in subtly different ways. This cannot really be captured by a translation. It is akin to the method new words are acquired by a native speaker, when that person starts to read widely.
Once a certain level is reached, a Latin-Latin dictionary should be used ( there are now a few online, none really geared for the student, though)
I think we have had a similar discussion here about types of ablatives a few months ago - most Romans would have been flummoxed by our explanations and categories - they just had internalised semantic webs for ablatives, that built up over time through huge amounts of exposure - and these neural webs were constantly growing.
Reading poetry expands them even further, as it stretches the natural language in unusual directions ( one reason why I am very much against novice students ( read, students in their first 5 or six years) reading complex Latin poetry such as Vergil, is that it totally messes up their mental map of natural Latin.)
To paraphrase the grammarian who says this the most clearly, J.W. Underwood - the key is much reading. Lots and lots of it. And even more. And, if you can get hold of audio, much audio ( most of us do not live in places where we can speak or hear actual people using Latin regularly). Audio is portable. Books require 'time out' . Even better than audio, would be video, and there is now a growing amount of that online, but once again, video requires one to sit and listen.....but audio is portable. And of course, if you have people to talk to, so much the better.