Prosodic Phrasing and Syllable Prominence in Spoken Prose. A Validated Coding Manual

Current systems for predicting prosodic prominence and boundaries in texts focus on syntax/semantic-based automatic decoding of sentences that need to be annotated syntactically (Atterer & Klein 2002; Windmann et al. 2011). However, to date, there is no phonetically validated replicable system for manually coding prosodic boundaries and syllable prominence in longer sentences or texts. Based on work in the fields of metrical phonology (Liberman & Prince 1977), phrase formation (Hayes 1989) and existing pause coding systems (Gee and Grosjean 1983), we developed a manual for coding prosodic boundaries (with 6 degrees of juncture) and syllable prominence (8 degrees). Three independent annotators applied the coding system to the beginning pages of four German novels and to four short stories (20 058 syllables, Fleiss kappa .82). For the phonetic validation, eight professional speakers read the excerpts of the novels aloud. We annotated the speech signal automatically with MAUS (Schiel 1999). Using PRAAT (Boersma & Weenink 2019), we extracted pitch, duration, and intensity for each syllable, as well as several phonetic parameters for pauses, and compared all measures obtained to the theoretically predicted levels of syllable prominence and prosodic boundary strength. The validation with the speech signal shows that our annotation system reliably predicts syllable prominence and prosodic boundaries. Since our annotation works with plain text, there are many potential applications of the coding system, covering research on prose rhythm, synthetic speech and (psycho)linguistic research on prosody.