Improving Communication of Visual Signals by Text-to-Speech Software

Abstract

Printed signals are well-documented aids to reading and memory for expository text. Despite their usefulness, many TTS applications fail to adequately communicate signaling. A theoretical framework called "SARA" provides an analysis of printed signals that identifies what specific information should be rendered in order to preserve the signaling function. Further, SARA identifies two important criteria -the availability criterion and the accessibility criterion -that should help guide the evaluation of alternative auditory renderings of signals.

Effects of audio headings on learning.

“…Finally, neither Experiment 1 nor Experiment 2 investigated the effectiveness of the typo-phonic rendering strategy (Maurel, 2004; Maurel et al, 2003; Lorch & Lemarié, 2013). A decision was made to exclude this type of rending strategy because previous findings showed its effects are similar to those of a prosodic rendering strategy (Lorch et al, 2016).…”

“…However, a prosodic rendering strategy might better support performance in a note-taking and cued recall task because the heading information (content and hierarchical level) is integral to the statement of the heading and thus might be more efficiently processed. In addition, a prosodic rendering strategy is most compatible with how people process speech naturally (Lorch & Lemarié, 2013). For example, people naturally change their pitch or volume if they want to emphasize a point while speaking.…”

“…A few researchers have proposed a guiding framework for translating visual information conveyed by signaling devices such as headings from print format into audio format (Maurel, 2004; Maurel, Lemarié, & Vigouroux, 2003; Lorch & Lemarié, 2013). The proposed model, MOTELS, stands for Modèle d ′ Oralisation des Textes Écrits pour être Lus Silencieusement (model of oralization of written texts for silent reading).…”

“…Under this constraint, note-takers are forced to be selective with the information that they have been given. Because audio headings are effective in signaling important topic information in an audio text (Lorch et al, 2012; Lorch & Lemarié, 2013), a note-taking task should be sensitive to listeners’ ability to capture important topics using audio headings.…”

“…The majority of efforts to improve TTS programs have been focused on making computer generated speech sound more human-like. Relatively little effort has been made to translate formatting information and information about emphasis in a printed text (Lorch, Chen, & Lemarié, 2012; Lorch & Lemarié, 2013). For example, the amount of white space between sections of text and indentations at the beginning of paragraphs communicates information about text structure.…”