Pavel Machač

Principles of Phonetic Segmentation

Within the speech sciences, consistency in speech segmentation is an issue, for example in corpora development and for reproducibility across studies. The objective of Principles of Phonetic Segmentation is to provide wellgrounded and consistent segmentation conventions. The book is targeted as a user-friendly reference for psycholinguists, phoneticians and speech technologists. In the first chapter the authors refer to inter-labeller discrepancies in previous literature, thereby defending their motivation to help reduce such discrepancies. This introduction is followed by definitions of what constitutes speech sounds and how boundaries between them can be conceptualized. The authors are clear that segment boundaries are to some extent arbitrary, however, necessary. The first chapter also includes explanations concerning language material (English and Czech, read and spontaneous speech), terminology and key reference points in the spectrograms. The segmentation conventions are described in chapters 2–12, which form the main part of the book. The descriptions are based on consonant boundaries and are organized according to (i) consonant types and (ii) whether the consonant is intervocalic or following/preceding other consonants. The main consonant types described are plosives (chapter 2), fricatives (chapter 3), nasals (chapter 4), trills (chapter 5), glides (chapter 6), lateral alveolar approximants (chapter 7), obstruents (chapters 8 and 9) and liquids (chapter 9). Vowels are treated as a homogeneous category for the purposes of this book. Chapters 2–7 focus on intervocalic (i.e. vowel-consonantvowel) segment boundaries, whereas chapters 8–10 focus on consonant-consonant boundaries. Chapter 11 deals with the segmentation of glottal stops preceding vowels, and chapter 12 addresses segment boundaries at the beginning and end of utterances. Each chapter is structured to first give an articulatory and acoustic introduction to the different sound groups, followed by a description of general segmentation rules, a set of less obvious examples, additional segmentation guidelines, and a summary. Throughout the text the descriptions are illustrated by spectrograms and waveforms, highlighting the cues and elements described. The final chapter 13 summarizes the book and offers evidence that the inter-labeller segmentation based on the conventions presented in this book gives a high level of consistency compared to previous work. The final chapter also gives information on which segment boundaries are drawn least consistently based on the same conventions. The book is in most respects user-friendly for the target audience, offering rigorous and sensible segmentation conventions. The most userfriendly aspect of the book lies in the extensive use of visual illustrations for the overall clearly written descriptions. Different examples are used to support the individual types of segment boundaries (e.g., plosive-vowel). Moreover, both canonical and less canonical examples are presented, which is likely to be useful for a user who, naturally, has to deal with a large amount of variation in his/her material. The book is also useful in that it gives a range of relevant tips on how to identify certain acoustic features in a spectrogram. These features include antiformants (in nasals and laterals) and typical formant values for different sound types as well as for transitions between them. As such, this book also works as a reference and guide for phonetic analyses, supplementing more extensive literature on this matter. The least user-friendly aspect of the book lies in its organization: It is not entirely transparent how the sound categories are defined and divided, and they are not indexed, which together might affect the user’s ability to locate descriptions of particular segment boundaries. At first glance, it appears that the chapters are organized according to the phonetic properties of the corresponding consonants, i.e. plosives, glides, etc. However at the same time the descriptions are organized according to some phonological association. As an example, taps/flaps are initially described under intervocalic plosives

On the impact of labialization contexts on unit selection speech synthesis

This paper presents a study on coarticulatory labialization and the significance of its respecting/violation during selection and concatenation of speech units in the unit selection speech synthesis. The aim of this study is to improve the overall speech quality, especially to increase the perceptual inconspicuousness between concatenated units. The labialization importance was verified by two listening tests-for phonetic laymen and specialists. To suppress the influence of other factors, both tests contained utterances with specially selected phones in specific contexts with respected and violated labialization. The preference for items with correct labialization was evident, which confirms the benefit of considering coarticulatory labialization in a unit selection speech synthesis.

Implications of Acoustic Variation for the Segmentation of the Czech Trill /r/

The Czech alveolar sonorant trill /r/, like liquids generally, constitutes a challenge from the point of view of locating its boundaries in the acoustic stream. As it is desirable to label and segment a phonetic corpus uniformly and also to facilitate a high degree of inter-labeller agreement, rules for specifying speechsound boundaries should be unambiguous and as straightforward as possible. In this study, we examined various acoustic forms of Czech /r/ --- from the trill and a flap to strongly reduced instances --- and their implications for segmentation. The above-mentioned requirements resulted in the necessity to treat the segmentation of intervocalic items of /r/ differently from items occurring in consonant clusters.