CasperSent: a program for computer-assisted speech perception testing and training at the sentence level.
The CasperSent program was designed mainly for auditory, visual, and auditoryvisual speech-perception training at the sentence level. Design is informed by a model of speech perception with 4 main components. Two are derived from the speech stimulus, namely sensory evidence and contextual evidence. The other 2 are provided by the perceiver, namely knowledge and skill. The principal target of training with CasperSent is perceptual skill. This involves such things as attention control, use and balance of sensory and contextual evidence, balance of speed and accuracy, confidence, risk-tolerance, and error-handling. The software may also be used for performance assessment while measuring the effects of talker, perceptual modality, angle of view (if lipreading), topic knowledge, sentence length, sentence type, and, if used in combination with a clinical audiometer, signal-to-noise ratio. The software operates on a personal computer with a minimum of peripheral equipment. Testing and training can be self-administered, helper-assisted, or clinician-controlled. Stimuli consist of 720 sentences. Performance measures are based on sets of 12 sentences, and are automatically logged for later recall, graphing, and analysis. It has been shown that performance improves over time, that rate of improvement depends on the type of feedback, and that prior knowledge of sentence topic increases performance. In addition to its intended application in aural rehabilitation, CasperSent also has potential value in research and teaching..