You're starting from the position that it's possible for a robot or other non-human to become completely native and fluent in a natural human language, even assuming anatomy permits the required articulation.
It doesn't appear to be even theoretically possible - because natural human language is infinitely combinatory, every speaker is capable of a completely original remark, and of interpreting every original remark in its context, which might include factors external to the conversation (relationships, cultural references, societal norms, past events, temper, tone, volume, etc). Although a robot can learn the phonetics, phonology, morphology, syntax and semantics of natural language, pragmatics does not appear to be achievable because it involves an understanding of the human experience, which we know to be personal and indescribable. And as for the higher apes, they don't have enough of the right kind of brain tissue for certain required operations such as theory of mind.
However, if you want the best possible then for a robot you'd look at a purely theoretical approach - like a university-level linguistics degree, including the acoustics of sound production; the rules of combination including word types, verb endings and plurals, word order, idioms, etc; a reasonably comprehensive lexicon...
... whereas for higher primates that is all impossible and you teach them as you teach a child, by repeated exposure. So far they appear to stall at the level of a toddler, where the other cognitive functions (referring to absent things for example, or projecting others' experiences and adjusting accordingly, etc) do not develop.
Some very interesting work exists here regarding the language of children with certain SN or brain damage, and the language deficit in adults following brain injury or degeneration.