Is there a way to get the best translation pair?

[Ughuuu]

Question

I know it's weird to define a "best" translation, but lets say the literal translation if not. Eg for word in English "agility" in Romanian (eg.) there are a few translations:

• agilitate f
• agerime f

I am a native Romanian speaker and I would prefer the translation agilitate in this case. The second is not as used (but it is correct). Is there any way to get the most accurate translations? (eg. similar in other languages).

My use case is I want to match the translations from 2 languages:

• eg. ro: agilitate
• eg. es: agilidad

These 2 match, but if it were another translation, it wouldn't be as good:

• eg. ro: agerime
• eg. es: agilidad

It gets even worse for other words, eg. abandon (in Romanian and Spanish), with multiple senses:

• ±to give up control of, surrender — Romanian: abandona (ro), părăsi (ro); Spanish: renunciar (es), suspender (es)
• ±to leave behind or desert; to forsake — Romanian: părăsi (ro), abandona (ro); Spanish: abandonar (es)

I am also interested mostly in the most popular sense, in this case for abandon I would like the translations to be:

• abandona (ro)
• abandonar (es)

Ideas

Some ideas I had for determining best translations were:

• for obvious words to check the root to see if it matches, but that wouldn't apply to all words.
• To check back and go from all words back to english one (if available in the dictionary). Eg. from agilitate -> agility?, and from agerime -> agility?. Problem is though that both of them translate back to agility.
• I though to look at the first translation, but it's not always the correct one.

For determining best sense, I though to either take first sense, or to take the sense with most translations.

[xxyzz]

• you can use translation data extracted from Spanish Wiktionary: https://kaikki.org/eswiktionary, but some words may not have Romanian translation
• maybe you can try some language models to compare the word and sense embedding