Spoken language interaction: Effects of vocabulary size, user experience, and expertise on user acceptance and performance

To enable novice and expert nursing staff to enter patient information into the healthcare database by speech, a prototype spoken language interface (SLI) has been designed. The prototype permits continuous spoken data to be entered using a speech recognition system. Feedback from the speech recognition system is provided by an auditory beep. The prototype interface contains two available vocabularies, a small specific vocabulary and a large inclusive vocabulary.;A split-plot experimental design tests the effects of the two levels of vocabulary available to the prototype spoken language interface, the two levels of nursing expertise (i.e., novice and expert) and two levels of experience with the interface. These three independent variables; vocabulary size, expertise, and experience, are investigated for their impact on the users performance and acceptance.;Each subject performs five cardiovascular assessments; a hands-busy and eyes busy task on a non-responding patient. Twenty one data items are gathered and entered into the spoken language system while performing the assessment. Performance is measured by time on task. Acceptance in obtained by the completion of a subjective acceptance questionnaire.;Three factors were found to affect the performance and acceptance of an SLI user. Increases in performance were found when the user was an expert in the domain, had experience with the SLI, or was using a SLI with a small vocabulary. A high degree of acceptance was found when the user was experienced with the SLI, was using a large vocabulary, or was a novice in the domain. No effects were found for comparisons of vocabulary and expertise or vocabulary and experience. The overall acceptance of the SLI was relatively high with mean acceptance percentage scores around 70%.