Health warning: use of speech synthesis can cause personality changes

Speech synthesis has been shown to offer high levels of intelligibility. Studies have however shown that the usability and naturalness of the speech quality of synthesized speech is less than optimal. This paper extends previous work on intelligibility and reports a pilot study to explore the perceived personality changes from a user point of view which are produced when speech synthesis is compared to natural speech. The paper describes an experiment involving 30 participants who listened twice to a short definition (4 or 5 sentences in length) of an obscure word using an internal telephone connection. After hearing each definition participants completed a personality questionnaire. One of the definitions was delivered using a (commercially available, 1998 vintage) speech synthesis system; the second definition used a recording of the voice of the professional recording artist on whose voice the synthesized speech was based. The order of presentation of the two voices was balanced across the participant sample. Personality profiles were constructed for both voices and statistical tests were used to identify the personality traits that differed significantly between the two voices. The paper includes experimental results, which demonstrate a range of perceived personality changes in the synthesised version of the actor's voice. The personality of the synthesised speech voice was characterised as someone who is introverted, conventional, neither agreeable or disagreeable but fairly conscientious, dependable, responsible and productive