Hux Karen, Knollman-Porter Kelly, Brown Jessica, Wallace Sarah E
University of Nebraska-Lincoln, United States.
Miami University, United States.
J Commun Disord. 2017 Sep;69:15-26. doi: 10.1016/j.jcomdis.2017.06.006. Epub 2017 Jun 23.
Using text-to-speech technology to provide simultaneous written and auditory content presentation may help compensate for chronic reading challenges if people with aphasia can understand synthetic speech output; however, inherent auditory comprehension challenges experienced by people with aphasia may make understanding synthetic speech difficult. This study's purpose was to compare the preferences and auditory comprehension accuracy of people with aphasia when listening to sentences generated with digitized natural speech, Alex synthetic speech (i.e., Macintosh platform), or David synthetic speech (i.e., Windows platform). The methodology required each of 20 participants with aphasia to select one of four images corresponding in meaning to each of 60 sentences comprising three stimulus sets. Results revealed significantly better accuracy given digitized natural speech than either synthetic speech option; however, individual participant performance analyses revealed three patterns: (a) comparable accuracy regardless of speech condition for 30% of participants, (b) comparable accuracy between digitized natural speech and one, but not both, synthetic speech option for 45% of participants, and (c) greater accuracy with digitized natural speech than with either synthetic speech option for remaining participants. Ranking and Likert-scale rating data revealed a preference for digitized natural speech and David synthetic speech over Alex synthetic speech. Results suggest many individuals with aphasia can comprehend synthetic speech options available on popular operating systems. Further examination of synthetic speech use to support reading comprehension through text-to-speech technology is thus warranted.
如果失语症患者能够理解合成语音输出，那么使用文本转语音技术同时提供书面和听觉内容呈现可能有助于弥补长期存在的阅读障碍；然而，失语症患者所经历的内在听觉理解障碍可能会使理解合成语音变得困难。本研究的目的是比较失语症患者在听由数字化自然语音、Alex合成语音（即苹果电脑平台）或David合成语音（即Windows平台）生成的句子时的偏好和听觉理解准确性。该方法要求20名失语症参与者中的每一位从与60个句子中的每一个相对应的四幅图像中选择一幅，这60个句子包括三个刺激集。结果显示，数字化自然语音的准确性明显高于任何一种合成语音选项；然而，对个体参与者表现的分析揭示了三种模式：（a）30%的参与者在任何语音条件下的准确性相当；（b）45%的参与者在数字化自然语音和一种（而非两种）合成语音选项之间的准确性相当；（c）其余参与者听数字化自然语音时的准确性高于听任何一种合成语音选项时的准确性。排名和李克特量表评分数据显示，与Alex合成语音相比，参与者更偏好数字化自然语音和David合成语音。结果表明，许多失语症患者能够理解流行操作系统上可用的合成语音选项。因此，有必要进一步研究通过文本转语音技术使用合成语音来支持阅读理解。
