Testing the effects of acoustic/prosodic entrainment on user behavior at the dialog-act level
Entrainment has been documented across several dimensions of human-human dialog. Experimental studies relating dialog success and acoustic/prosodic (a/p) entrainment in spoken dialog systems, point towards a non-neutral effect of a/p entrainment. But results also suggest the presence of positive eff...
Guardado en:
| Autores principales: | , , , , , , , , , |
|---|---|
| Formato: | CONF |
| Materias: | |
| Acceso en línea: | http://hdl.handle.net/20.500.12110/paper_23332042_v2018-June_n_p374_Gauder |
| Aporte de: |
| Sumario: | Entrainment has been documented across several dimensions of human-human dialog. Experimental studies relating dialog success and acoustic/prosodic (a/p) entrainment in spoken dialog systems, point towards a non-neutral effect of a/p entrainment. But results also suggest the presence of positive effects of disentrainment. A plausible driver behind this last result could be that in these experiments both users and avatars were restricted in their use of dialog acts. In particular, systems only produced answers which entrained to the sole dialog act produced by users: requests for advice. Given that there is significant documented correlation between dialog acts and a/p features, it seems reasonable to hypothesize that a/p entrainment may occur at the dialog-act level and that entraining or disentraining across dialog acts may introduce misleading artifacts. This paper presents the design and implementation of an experimental setup which allows to implement entrainment and disentrainment policies at the dialog-act level. It also presents results of a pilot study in Argentine Spanish. © 2018, International Speech Communications Association. All Rights Reserved. |
|---|