While I appreciate that I can use SSML to correct the pronunciation of specific words, the functionality is cumbersome when dealing with topics that use such words repeatedly. For example I have been working on a "Yoga Facts" app on an off for about a month. Much of the language in yoga is based on Sanskrit so these words inevitably require tweaking with SSML tags. For example, the word asana (a yoga pose) needs to be output as <phoneme alphabet="ipa" ph="ˈɑːsənə">Asana</phoneme> to sound right. I used to work with the Microsoft Speech API. MS allowed developers to create custom "Lexicons" to provide custom word pronunciations during speech synthesis. Under this scheme I would provide a custom lexicon for my application that included the correct pronunciation for asana and then every time the speech engine ran into the work asana, it would speak it correctly. Anyway, it would be very cool if the AVS had something similar.