Effects of sentence structure and word complexity on intelligibility in machine-to-human communications. (November 2019)

• Record Type: Journal Article
• Authors: Hui, C.T. Justine; Jain, Sahil; Watson, Catherine I.
• Is Part Of: Computer speech & language. Volume 58(2019)
• Page Start: 203
• Page End: 215
• Publication Date: 2019-11
• Subjects: Dialogue design -- Speech perception -- Speech synthesis -- Machine to human communications; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2019.03.002
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: Complex syntax and simple words contribute to misunderstandings in listening to unfamiliar synthetic speech. Created phonetic complexity measure for perception from production based measures. Effect of how phonetically distinctive the word is is greater than the effect from word occurring frequency in the language on word comprehension. Significant difference between native and non-native speakers on synthetic speech understanding. Abstract: With the rise of robotics and artificial intelligence, good communication between humans and machines becomes more important. However, users with language and hearing disadvantages may find synthetic speech systems to be difficult to understand. In this study, we explore the types of sentence structure and level of word complexity that affect intelligibility of speech in unfamiliar context. Using semantically unpredictable sentences, we found that sentence with more complex syntax such as relative pronouns and question words are harder to comprehend, while on the word level, it is the shorter and simpler words that contribute to misunderstandings. We found that although word frequency affects how well a word is recognised, the effect from the occurring frequency is much less than the effect of how phonetically distinctive the word is. There was also evidence of significant difference between native speakers and non-native speakers on how well they could understand the sentences. These results may help us in designing better dialogue system for machine to human interactions, especially in the healthcare arena, where often users have disadvantages in language and hearing abilities.