How far can prosodic cues help in word segmentation?

Prosodic cues are of great importance in parsing speech signal into prosodic and lexical units. Listeners detect the changes of the prosodic parameters and interpret them to detect sentence modalities or the mood of the speaker. Some automatic speech recognition systems try to use prosodic parameters to detect boundaries of prosodic units and help thus the acoustic decoding process. Although the automatic detection of major prosodic boundaries is most of the time reliable, minor boundary detections are prone to error. However, listeners, unlike automatic processing systems, can detect with great precision boundaries of lexical items even if they do not coincide with major prosodic boundaries. Our feeling is that a deeper understanding of the prosodic parameters in spontaneous speech would improve their modeling and ultimately their use by automatic systems. This study analyses filled and silent pause occurrences and two prosodic parameters, duration of pauses and vowels and F0 slopes, measured on a spontaneous speech corpus in French. The results of the analysis revealed that a simple local comparison of the parameter values with the values measured in the vicinity of the segment under consideration can provide valuable information on the lexical boundaries as well as on prosodic patterns of the lexical units.