Generation and perception of f_0 markedness in conversational speech with adverbs expressing degrees

Aiming at natural F0 control for conversational speech synthesis, F0 characteristics are analyzed from both generation and perception viewpoints. By systematically designing conversational situations and utterances with adverb phrases expressing different degree of markedness, their F0 characteristics are compared. The comparison shows the consistent F0 control dependencies not only on adverbs themselves but also on the attribute of neighboring adjective phrases. Strong positive/negative correlation is observed between the markedness of adverbs and F0 height when an adjective phrase with a positive/negative image is followed to the current adverb phrase. These consistencies have been perceptually confirmed by naturalness evaluation tests using the same two-phrase samples with different F0 heights. These results indicate the possibility of F0 control for natural conversational speech using lexical markdness information and adjacent word attributes.