This article investigates whether authentic native speaker (NS) to NS speech can be made available to the learner listener through the use of a novel slow-down tool. Results from various preliminary tests seem to indicate that the use of a slow-down algorithm in many cases, and in particular in samples with a higher speed rate and word count, leads to an improvement in subjects’ ability to perceive and understand what was being uttered in the samples. Tests revealed that even NS listeners, as opposed to non-native (NN) listeners, prefer to hear authentic NS speech which is either unscripted or is influenced by regional accent, at a slowed down speed. It also seems that ‘unexpected’ words (such as words with high contextual value, but which cannot be processed in a top-down fashion because of the size of the sound snippet) are initially not understood at the original speed of delivery, even in a scripted and carefully pronounced pedagogic sample. Samples containing chunks or formulaic sequences, however, appear to be easily understood at 100% by the majority of NS listeners due to the holistic processing of these language units.

No abstract

A common suggested treatment for verbal apraxia is repetition, and the use of slow speech. The required slow speech may be attained by time-scaling ordinary-speed speech. However, when used for this purpose, the quality of the expanded speech must be of a very high quality to be of pedagogical benefit. This paper describes a new method of time-scaling based on the knowledge of speech characteristics, the relative durations of speech segments, and the variation of these durations with speaking rate. The new method achieves a high quality output making it suitable for use as a computer-assisted speech therapy tool.

No abstract

No abstract