I am playing with the idea of building a language parser/transformer and getting into the structure of pronunciations. This leads me to try and come up with rules for the parser for understanding how to transcribe a language from letters in the source language to romanized output. From what it sounds like, rule-based systems are what have been around for a while, but there is new research on constraint-based processing of the pronunciations. Wondering what are some good resources to get up to speed.
The best entry-level introduction I've found is Matt Gurevitch's OT Machine, a tool for conlangers to build and test out Optimality Theory-based models. It includes an overview of how OT works in general which is significantly more accessible than most published papers on OT. (Optimality theory has problems with opacity, in more ways than one!)
The original paper which spawned the whole theory is Prince and Smolensky 1993, but I would absolutely not recommend that for someone starting out unless they already have years of background in computational linguistics.
There is some confusion in the question owing to the use of the word "pronunciation". There is a constraint-based theory of phonological computation, Optimality Theory, which is the dominant theory in phonology. However, it is not about "pronunciation", it is about phonemic outputs qua sequences of letters (in IPA or some other transcriptional system). If you can go from [tadˁaʕu] to a pronunciation and don't care about the particular details of the language (it's Classical Arabic), then OT could be relevant. An OT account might tell you how to describe the mapping from /tawadˁiʕu/ to [tadˁaʕu].
Paul Boersma, however, is into the physical output stuff and he seems to do constraints. From a phonetic perspective, I'd recommend starting there. If you want to learn OT as a phonological theory, you could read R. Kager Optimality theory and J. McCarthy Doing Optimality Theory: Applying theory to data.