摘要 |
An automated prosodic boundary markup mechanism is implemented in a text-to-speech converter system. The mechanism parses text at 116 and assigns prosodic boundaries to the text at 118 based on the parse. However, this results in overmarking of prosodic boundaries. Prosodic boundary constraints are applied to the marked up text based on the parts of speech assigned at 114 to the words in the text portion to remove inappropriate boundaries. For example, a prosodic boundary is removed if it immediately follows a deaccented word, if there is only one word between the boundary and punctuation, if the boundary lies between two words of the same syntactic class, or if the boundary immediately follows a function word. |