Memory, comprehension, and judged quality of synthetic speech under several prosodic conditions
Astrid McHugh, Leah S. Larkey · The Journal of the Acoustical Society of America · 1977
Prosodic information contributes to the understanding of natural speech. Stress and timing convey semantic and syntactic information. Similarly, acceptable synthetic speech depends on more than correct phonemic specification. Appropriate intonation and timing patterns should improve the speech quality. Not only should such speech be judged to sound better, but one might also expect improvement on a number of other tasks such as memory, comprehension, or listening under adverse conditions (e. g., noise). Tests were conducted using multiple versions of synthesized sentences, paragraphs, and word lists. All had the same phonemic sequence but varied in prosodic specification. Possible versions were as follows: no stress information (a monotone), inflection information alone (pitch/loudness changes), timing information alone, or both inflection and timing. Listener ratings were very consistent: the more correct prosodic information was available, the better the speech was judged to sound. Scores on a simple comprehension test were higher for “good” versions only with unpracticed listeners. There was some evidence that with a more stringent memory test, scores were higher for prosodically good versions. [Work supported by NINDS, NICHD, and NIMH.]