Learning Phonotactics from Linguistic Informants