Title :
Shaping to convergence: experiments with speech graffiti
Author :
Tomko, S.L. ; Rosenfeld, R.
Author_Institution :
Sch. of Comput. Sci., Carnegie Mellon Univ., Pittsburgh, PA
Abstract :
Various analyses have shown that conversation participants converge in a variety of ways in both human-human and human-computer interactions. This finding has potential applicability in spoken dialog systems, as users could be encouraged to converge to vocabulary or syntax preferred by the system. This paper reports on a user study investigating three versions of system prompts designed to promote convergence to the Speech Graffiti interaction protocol. We found that all three versions engendered significant convergence over time, although the versions varied in their immediate shaping power. We also found that gender and initial grammaticality had strong effects on overall performance.
Keywords :
convergence of numerical methods; human computer interaction; human factors; interactive systems; protocols; speech recognition; human factors; human-computer interaction; human-human interaction; shaping power; speech graffiti interaction protocol; speech recognition; spoken dialog systems; user interfaces; Computer science; Convergence; Human factors; Natural languages; Protocols; Speech analysis; Speech processing; Speech recognition; User interfaces; Vocabulary;
Conference_Titel :
Spoken Language Technology Workshop, 2006. IEEE
Conference_Location :
Palm Beach
Print_ISBN :
1-4244-0872-5
DOI :
10.1109/SLT.2006.326840