Analiza wybranych cech zmienności iloczasowej w różnych stylach wypowiedzi na podstawie korpusów nagrań dla technologii mowy
Analysis of Selected Aspects of Durational Variability in Different Speaking Styles Based on Speech Technology Corpora

Summary/Abstract: This paper is aimed at describing selected durational and rhythmic features of contemporary Polish based on automatic annotation mining and analysis of recordings of speech representing varying speaking styles. The study is based on corpus data compiled or created within automatic speech recognition projects for Polish. The corpora were recorded in different types of field conditions (e.g., parliamentary speeches, authentic spontaneous telephone conversations, police scene view reports), as well as in laboratory environment (read speech and task-oriented dialogues). Utterances representing different speaking styles are not only characterized by different syntactic or lexical features but more importantly—they sound different. The paper reports on the results of an investigation into timing variability in utterances depending on the speaking style, i.e. syllable durations, their mutual relationships in neighbouring contexts, local and global speech rate variability. Several existing methods of speech timing and rhythm representation were used to analyse patterns based on syllable durations values and relationships: Time Group Analysis (Gibbon 2013), nPVI (Grabe, Low 2002), Duration Quadrants (Wagner 2007). All measurements were conducted using Annotation Pro freeware tool, and its plug-ins (Klessa 2016a).

• Details
• Contents