I have been working on an algorithm to combine names with pronunciations that overlap, based on ARPAbet phoneme representations from CMU dict + generated by LLM

I have been working on an algorithm to combine names with pronunciations that overlap, based on ARPAbet phoneme representations from CMU dict + generated by LLMs. The data is still messy, because "creative" spellings are hard for LLMs to deal with: token overlap with primary names is frequently low, and spelling variations only exist because English has such a lax approach to phonics to begin with. Here's Ryan with its spelling variations: https://nameplay.org/names/combined/popularity/Ryan https://nameplay.org/names/combined/popularity/Ryan