Multi‐language speech synthesis—A search for synthesis universals

In the past few months, Cornell's interactive speech synthesis system has been used to develop synthesis rules for a variety of languages (German, Spanish, and Indonesian). In each case, intelligible speech was produced after only a few weeks of work. This rapid development can be attributed to two factors. The first is the flexibility of the synthesis system itself, which allows users to easily test competing synthesis strategies. The second is the overlap between our English synthesis rules, which existed when the multi‐language work began, and the rules in the other languages. This overlap suggests that some of these rules may be appropriate for many or even all languages. The use of the Cornell Speech Research System to develop rules for more languages should shed light on the following questions: (1) To what extent do the synthesis rule sets for different languages overlap? (2) How can this overlap be explained ? (3) What implications will these observations have for phonological theory? Will they he...