Note that Frank Liang's paper is on hyphenation, NOT on syllable detection. In addition, his thesis paper itself states that its success rate is around 89% for the dictionary he used, which is not going to be good enough for everyone. There really is no substitute for manually doing it for every single word it seems. It's not that efficient to have to require a complete one-to-one lookup table wordlist in order to do it, but these days storage space is far less expensive than CPU time anyways.
Perhaps someone might consider making a CAPTCHA-like service so that many users could be asked to provide the solution to every known word, with the results checked against each other, so that one person wouldn't have to do it all themselves. I'd hope the results would be released freely once complete.
