Welcome to The Protag Project Website

Abstract

In linguistics, variable pauses and fixations during reading have been linked to prosody. Prosody refers to properties of speech which add communicative meanings to spoken language. Prosody in reading is reported to influence the processing of the syntactic structure of a string of words in a similar way it supplements the expression of spoken language. In a previous research [4] , Marks, Keeling, Anderson and Trimble (2009) produced PEDText, a dynamic text presentation technique which sought to sustain comprehension at higher rates of reading through the projection of visual prosodic properties derived from plain, electronic text. Results were promising; however, PEDText relied on a manual method of prosodic annotations on plain text which theoretically affected accuracy as well as the practicality of handling large volumes of text. This study explores an automated method for annotating plain electronic text with prosodic indicators in order to address the highlighted limitations of PEDText. The implemented solution, ProTag, utilizes a prosodic annotated corpus of spoken language (ProPOSEC [1] – Atwell and Brierley, 2010) to achieve automated prosodic tagging. ProTag produces results that are not significantly different from the results obtained through PEDText’s existing manual method. When compared to human speech, however, ProTag was found to be less accurate than manual methods. The results indicate that the use of an automated prosodic tagging engine, with further refinement, may be a viable means of automating the annotation of prosodic properties on plain electronic text.