Role of strength relations in the patterning of segmental speech sounds: evidence from Indian languages