Play it again: a study of the factors underlying speech browsing behavior

Abstract

Several recent UIs support access to recorded speech archives, but these have not yet been systematically evaluated. We describe a laboratory study of speech archive browsing using a GUI. We evaluate the effects of four factors: task type, familiarity, structure, and play operation duration. We found that while users learnt the overall layout of topics in the archive, they experienced major problems in learning the internal structure of archival topics. Contrary to our expectations, we also discovered that structural information and fixed duration play operations were less useful for browsing than anticipated. We discuss the impact of our results for speech archive UI design, and describe a new UI which supports navigation within topic.