ISCA & IEEE Workshop on Spontaneous Speech Processing and Recognition April 13-16, 2003 Tokyo Institute of Technology, Tokyo, Japan

Naturalness in Speech Communications

Paul C. Lustgarten (1), B. H. Juang (2)

(1) Avaya Labs Research, Basking Ridge, New Jersey, USA
(2) Avaya Labs Research and Georgia Institute of Technology, Atlanta, GA, USA

Speech has long been considered the most natural form of human communications. People with normal speaking and hearing abilities use speech to exchange information every day, often effortlessly. An interesting question, however, remains rarely addressed: What constitutes naturalness in speech communications and how is naturalness achieved? In this paper, we attempt to analyze the "human behavioral components" that contribute to the naturalness or perceived naturalness in human speech communications. We further attempt to mechanize one of these components, namely use of "reference", in an intelligent service scenario involving spoken language, to demonstrate that indeed such a component greatly adds to the natural experience in human-machine dialog. We hope to inspire further research into other naturalness dimensions that may enrich human machine interaction.

Full Paper

Bibliographic reference.  Lustgarten, Paul C. / Juang, B. H. (2003): "Naturalness in speech communications", in SSPR-2003, paper WMO2.