The Speech2IPA leaderboard currently only supports models that output IPA. There are a number of other alphabets that can be used to annotate phonemes, e.g., ARPAbet, and a number of models that output these alternative symbols (check their vocab/tokenizer), e.g., this model. We have code to convert between a number of these. We just need to add an option to select from these when submitting a model to the leaderboard. Bonus points for auto-detecting the symbolset and allowing entering a custom vocab to IPA mapping when submitting.
The Speech2IPA leaderboard currently only supports models that output IPA. There are a number of other alphabets that can be used to annotate phonemes, e.g., ARPAbet, and a number of models that output these alternative symbols (check their vocab/tokenizer), e.g., this model. We have code to convert between a number of these. We just need to add an option to select from these when submitting a model to the leaderboard. Bonus points for auto-detecting the symbolset and allowing entering a custom vocab to IPA mapping when submitting.