Russian Vowels Research Articles

Source-Oriented Generalizations as Grammar Inference in Russian Vowel Deletion

Speakers learn detailed generalizations about the morphophonology of their language and extend them to nonce words. We propose a theory of this morphophonological knowledge that partitions the lexicon into uniform and productive sublexicons. Each sublexicon has its own phonotactic grammar, which the speaker uses as an inference mechanism to determine the relative productivity of each sublexicon. We report the results of an experiment on the generalization of mid vowel deletion (‘‘yer’’ deletion) in Russian, showing that speakers encode source-oriented generalizations about the shapes of words that can undergo vowel deletion, as well as product-oriented generalizations about words that result from vowel deletion. An implementation of our model learns the patterns of deletion and captures both source-oriented and product-oriented generalizations.

The interaction of tone and vowel quality in Optimality Theory: A study of Moscow Russian vowel reduction

Although the interrelation of tone and segmental quality is typologically unusual, cases of vowel-tone interaction have been reported in the literature. The present paper argues that tone can interact directly with vowel quality without mediating factors such as syllable structure or duration. The basic assumption is that tonally prominent units co-occur with prominent segments. In terms of Optimality Theory, this generalisation is expressed by the family of markedness constraints *H/≤a, which are derived by harmonic alignment of two natural linguistic scales, the tonal scale and the sonority scale. The proposed constraints are used in the analysis of vowel neutralisations found in immediately pretonic positions in the Moscow variety of Standard Russian. A characteristic trait of standard Russian is that it exhibits two-pattern vowel reduction: moderate reduction is found in immediately pretonic positions, while extreme reduction is found in atonic positions. Previous accounts have suggested that different degrees of reduction are due to iambic foot structure. This paper argues that such an approach is flawed and develops an OT analysis, which is based on the insight originally expressed by Bethin (2006) that vocalic neutralisations are driven by the High tone spread from the stressed syllable.

Количественная редукция гласных на месте графических «а» и «о» после твердых согласных в русской речи в Литве

This article analyses the absolute duration (ms) of stressed Russian vowels /a/, /o/ (graphs: “a”, “o”) and their allophones in unstressed positions after the hard consonants in the pronunciation of native and non-native Russian speakers in Lithuania. The results of the conducted spectral analysis reveal the specificities of quantitative reduction in the speech of the Russian speakers in Lithuania and the Lithuanian speakers that are learning the Russian language. These specificities are influenced by the two phonetic systems interaction. The speakers of both languages by the realisation of “a” and “o” violates the relation of unstressed vowel duration that is peculiar to the contemporary Russian language: the post-stressed vowels in closed syllables are shorter than the pre-stressed vowels; the first pre-stressed syllable differs from the second pre-stressed and post-stressed syllables by a longer voice duration. Both Russians and Lithuanians pronounce vowels longer in post-stressed syllables than in the pre-stressed syllables. This corresponds to the qualitative reduction of the Lithuanian language vowels /a:/ and /o:/. There are certain differences between the pronunciation of qualitative vowels “a” and “o” reduction among the native and non-native Russian speakers in Lithuania. The Russian speakers in Lithuania pronounce the second pre-stressed vowel longer than the first pre-stressed vowel; this corresponds to the degree of reduction of pre-stressed vowels “a” and “o” in the standardised Russian language. These degrees of quantitative reduction in the Lithuanian pronunciation are peculiar only for “a” in the Russian language. According to the duration ratio, the unstressed allophones “a” and “o” in the Russian language are closer to the unstressed /a:/ and /o:/ in the Lithuanian language in the pronunciation of Russian-Lithuanian bilinguals than in the pronunciation Lithuanian speakers.

Differential Thresholds of Perception of Sound Duration in Different Phonetic Contexts

The paper presents the results of the study which was aimed at obtaining the values of differential thresholds for Russian stressed vowels. Two types of psychoacoustic experiments were conducted. The experimental material was formed taking into account the segmental and prosodic context. The segmental context of the vowel stimuli was relatively similar while prosodic context varied. Each vowel was presented in several word types which involved one-syllable words and three-syllable words with different stress patterns. Each word was embedded into an affirmative utterance (initial/medial/final position), an interrogative utterance (final position), and an utterance with a contrastive stress on the word. The subject listened to a sample word containing a target vowel and its modifications with increased or decreased vowel duration. The gradual increment or decrement was 1 F0 period or 10 ms on average. The task was to click the modification when the duration difference was perceived. Then the differential thresholds are calculated by means of statistical analysis. The average differential threshold was found to be 2–4 F0 periods which is approximately about 20–40 ms. The threshold value does not seem to be influenced by word duration (in syllables) or stress pattern.

Acoustic vocal tract model of one-year-old children

The physical shape of vocal tract and its formant (resonant) frequencies are directly related. The study of this functional connectivity is essential in speech therapy practice with children. Most of the perceived children's speech anomalies can be explained on a physical level: malfunctioning movement of articulation organs. The current problem is that there is no enough data on the anatomical shape of children's vocal tract to create its acoustic model. Classical techniques for vocal tract shape imaging (X-ray, magnetic resonance, etc.) are not appropriate for children. One possibility is to start from the shape of the adult vocal tract and correct it based on anatomical, morphological and articulatory differences between children and adults. This paper presents a method for vocal tract shape estimation of the child aged one year. The initial shapes of the vocal tract refer to the Russian vowels spoken by an adult male. All the relevant anatomical and articulation parameters, that influence the formant frequencies, are analyzed. Finally, the hypothetical configurations of the children's vocal tract, for the five vowels, are presented.

About neural-network algorithms application in viseme classification problem with face video in audiovisual speech recognition systems

The paper considers the phoneme recognition by facial expressions of a speaker in voice-activated control systems. We have developed a neural network recognition algorithm by using the phonetic words decoding method and the requirement for isolated syllable pronunciation of voice commands. The paper presents the experimental results of viseme (facial and lip position corresponding to a particular phoneme) classification of Russian vowels. We show the dependence of the classification accuracy on the used classifier (multilayer feed-forward network, support vector machine, k-nearest neighbor method), image features (histogram of oriented gradients, eigenvectors, SURF local descriptors) and the type of camera (built-in or Kinect one). The best accuracy of speaker-dependent recognition is shown to be 85% for a built-in camera and 96% for Kinect depth maps when the classification is performed with the histogram of oriented gradients and the support vector machine.

Quantitative Reduction of Vowels Following Soft Consonants in the Russian Speech of Lithuanians

This investigation of the acoustic characteristics of phonetic units in the Russian speech of Lithuanians is important for a description of the mechanisms of speech production in a situation of the interference of two languages (native and studied), and for the solution of applied problems of teaching (or correcting) Russian phonetics to Lithuanians.The article analyses the average duration (in milliseconds) of the Russian stressed vowels [i], [e], [a], [u], and their unstressed allophones, when they follow soft consonants in the speech of Lithuanians. The results of the spectral analysis show peculiarities of quantitative reduction of Russian vowels in the speech of Lithuanians, due to the interference of the phonetic systems of the two languages.Four major types of distinctions – in the phonological and phonetic systems of the genetically related Russian and Lithuanian languages, in the degree of quantitative reduction of vowels, in the relation between the duration of different types of unstressed vowels in different positions, and in the relation between the duration of unstressed vowels depending on the placement of the stress – allow deviations from Russian norms of pronunciation to be predicted in the Russian speech of Lithuanians.In unstressed syllables, Lithuanians pronounce vowels of varying duration in place of the graphemes и, е and я, which testifies to the unequal degree of quantitative reduction of these vowels. In Russian, these vowels in unstressed positions are expressed by the same sound and cease to differ. In the pronunciation of Lithuanians, the weak quantitative reduction of the Russian vowel [a] (grapheme я) and the insignificant shortening of the duration of [u] can be observed in all unstressed syllables.According to the received data, in the Russian speech of Lithuanians, a poststressed, non-final vowel can be longer than a vowel in the second prestressed syllable; this breaks the opposition of duration of poststressed vowels to that of all pre-stressed vowels which is characteristic of Russian.

On paradigm uniformity and contrast in Russian vowel reduction

Reduction of unstressed /o/ and /a/ to [ɐ] or [ə] after non-palatalized consonants and to [ɪ] after palatalized ones in Contemporary Standard Russian (CSR) is systematic. But in certain inflectional suffixes [ə] occurs instead of the expected [ɪ] after palatalized consonants. In order to explain these apparent exceptions, I argue that vowel reduction after palatalized consonants is constrained by the morphology and that reduction to [ɪ] is blocked in certain cases of the paradigm by the interaction of Relativized Paradigm Uniformity (PU) and Paradigm Contrast (PCON) constraints (Rebrus and Torkenczy 2005; Steriade 2000). The main finding is that there is a critical contrast between the singular and plural within a given case, Number × Case, and [ɪ] is blocked when it would result in homophony with an /i/ [ɪ] suffix in the relevant situation. When the morphological contrast is implemented by some other means, then regular vowel reduction to [ɪ] takes place. The gen sg /-a/ suffix has special status due to type and token frequency of the [ə] variant. An analysis in terms of paradigm uniformity and contrast provides a more coherent account of the direction of language change in CSR vowel reduction than do appeals to stressed vowel faithfulness, spelling pronunciation, grammatical analogy or paradigm uniformity alone.

Perceptual assimilation of American English vowels by inexperienced Russian listeners

In a perceptual assimilation task, 16 Russian speakers categorized American English (AE) vowels [i[see text], I, epsilon, ae[see tedxt], a[see text], u[see text]] in /Vp/ disyllables into Russian (RU) response categories and rated their perceived goodness on a 9-point Likert scale. Cross-language discriminant analysis established acoustic similarities to Russian vowels. For all but AE [epsilon], acoustic similarity predicted modal assimilation responses. Russian listeners consistently assimilated AE point vowels to their Russian counterparts, whereas assimilation of the remaining vowels was less consistent. These acoustic and perceptual similarity patterns provide a baseline for future studies of AE vowel discrimination by Russian learners of English.

Russian Vowel Reduction and Phonological Opacity

The aim of this paper is twofold. First, it proposes a novel analysis of Vowel Reduction in Russian within the framework of Optimality Theory. Secondly, it discusses the opaque cases that result from the interaction of Vowel Reduction with other phonological processes, such as Glide Deletion and Retraction. The article demonstrates that Standard Optimality Theory, which evaluates output forms using one set of ranked constraints, cannot handle the Russian facts. It is argued that a modified version of Optimality Theory, Derivational Optimality Theory, can provide a straightforward account of the phenomena under consideration.

Speech Acoustics and Phonetics: Selected Writings (review)

Reviewed by: Speech acoustics and phonetics: Selected writings by Gunnar Fant Peter Ladefoged Speech acoustics and phonetics: Selected writings. By Gunnar Fant. (Text, speech and language technology 24.) Dordrecht: Kluwer, 2006. Pp. xiv, 322. ISBN 1402027893. $64.95. Gunnar Fant was the first of the great engineers to lay out the field of acoustic phonetics, and this new collection of his papers further documents his achievements. There is no doubt that he is a towering figure in the field. His Acoustic theory of speech production (Fant 1960) provided the solid foundation on which much later work was based. The nineteen papers that he has selected for this book cover a wide range of topics and make available several papers that are not easy to find. Nine of them appeared previously only in the quarterly reports of the Department of Speech, Music, and Hearing or other in-house publications of the Stockholm Royal Institute of Technology. Two of them have not been in print before, and several of the others are in conference proceedings, handbooks, or encyclopedias rather than readily available journals. The papers are grouped into six sections, each with a short introduction. The first section consists of a single paper, not previously in print, in which F gives an overview of his more than fifty years of speech research. F was born in October 1919 and began publishing papers on the sounds of Swedish in 1948. Since then he has been a very productive scholar; there are 247 publications listed at the end of this book. The second section, ‘Speech production and synthesis’, consists of five papers, starting with his early work at Ericsson Telephone Company, in which he lays out the mathematical foundations for the relations between formant frequencies, amplitudes, phase, and bandwidth. Other papers in this section include F’s account of the relation between area functions and the acoustic signal (one of the few publications in a refereed journal, Phonetica) and two papers on speech synthesis, one based on an articulatory model of the vocal tract, and the other, which is of considerable historic interest, describing the OVE II parametric formant synthesis. There is also a paper, published in 2001, updating his analyses of Swedish vowels, relating a three-parameter model of the vocal tract to the corresponding formant data. There is no doubt that this paper provides an interesting account of one way of parameterizing the vocal-tract shapes of a single speaker. But for a paper revised in 2001 it slips up by not updating references to other work on the same topic. It claims, for example, that there is ‘rather meagre data available on vowel specific VT area functions, which has long been dominated by the Russian vowels, (Fant 1960)’ (58). But by 2001 Baer et al. 1991 and other imaging studies were available. The third section has three papers on the voice source, all coauthored with Qigang Lin and the first also coauthored with Johan Liljencrants. These three papers describe a way of parameterizing the shape of the wave produced at the glottis, the source in F’s source-filter approach to the acoustics of speech. This parameterization, which is known as the L-F (Liljencrants-Fant) model, has become widely used in the literature on phonation types and voice quality. The fourth section, ‘Speech analysis and features’, starts with a paper published over forty years ago on ‘Descriptive analysis of the acoustic analysis of speech’. F presents a list of phonetic characteristics and their acoustic correlates. Emphasis is given to the fact that every segment is influenced by its neighbors. This is still a good, concise account of the information that can be obtained from spectrograms. The next paper in this section, which was published in 1997 in an encyclopedia of acoustics, concentrates on analysis and processing techniques. It provides more technical detail than many linguists will want, but is full of valuable information on methods of spectral analysis and extraction of formant frequencies. The third paper, ‘Features: Fiction and facts’, originally appeared in the proceedings of a symposium in 1986. It provides F’s later views on distinctive features. In the original publication [End Page 189...

Dynamic center of gravity effects: Matching virtual frequency glides

Lublinskaja (1996) demonstrated that changing the amplitudes of two‐resonance Russian vowels over time leads listeners to hear a glidelike transition that follows the dynamic center of gravity (COG). Anantharaman (1998) investigated COG transitions for sinusoidal signals by asking subjects to match frequency‐modulated (FM) to virtual frequency (VF) glides. The amplitudes of two tones were simultaneously modulated to produce a percept of a glide with rising pitch, similar to that of an FM sweep, although the frequencies remained fixed. However, listeners might perform Anantharaman’s task by simply attending to the location of one of the anchor tones without perceiving the virtual glide, as the frequencies of the two component tones were changed to alter the frequency span of the VF glide. This study modified that work by fixing the anchor frequencies, but confirmed the potential for procedural artifacts. Two additional experiments eliminated FM signals altogether and asked listeners to match VF glides to VF glides. Even with anchor frequency roving, subjects were able to match the VF glide trajectories for frequency separations up to 6 to 8 ERB. Results were compared with predictions from the perceptual spectral centroid model. [Research supported by a grant from NIH/NIDCD R01‐DC006879.]

Temporal and spectral variability of vowels within and across languages with small vowel inventories: Russian, Japanese, and Spanish

Variability of vowels in three languages with small vowel inventories (Russian, Japanese, and Spanish) was explored. Three male speakers of each language produced vowels in two-syllable nonsense words (VCa) in isolation and three-syllable nonsense words (gaC1VC2a) embedded within carrier sentences in three contexts: bilabial stops in normal rate sentences and alveolar stops in both normal and rapid rate sentences. Dependent variables were syllable duration and formant frequency at syllable midpoint. Results showed very little variation across consonant and rate conditions in formants for /i/ in Russian and Japanese. Japanese short /u, o, a/ showed fronting (F2 increases) in alveolar context, which was more pronounced in rapid sentences. Fronting of Japanese long vowels was less pronounced. Japanese long/short vowel ratios varied with speaking style (isolation versus sentences) and speaking rate. All Russian vowels except /i/ were fronted in alveolar context, but showed little change in either spectrum or duration with speaking rate. Spanish showed a strong effect of consonantal context: front vowels were backed in bilabial context and back vowels were fronted in alveolar context, also more pronounced in rapid sentences. Results will be compared to female productions of the same languages, as well as American English production patterns.

The unnatural /C ju/ (< foreign /Cy/) sequence in Russian loanwords: A problem for the perceptual view

LaCharité and Paradis (2000, 2005) and Paradis and LaCharité (1997, 2001, in press), who support the “phonological view”, have already brought numerous arguments against the “perceptual view” as a general explanation for sound adaptation in lexical loanwords. This article offers a new type of evidence against the perceptual view: the fact that, for phonological reasons, the adaptation of a phoneme of the source language yields in the borrowing language (L1) an unnatural phonotactic sequence. This sequence is /C ju/, where a consonant is palatalized before a back vowel (/u/). Yet palatalization systematically occurs before front vowels in Russian, not back vowels. The /C ju/ sequence in Russian results overwhelmingly from the adaptation of the foreign sequence /Cy/ in loanwords from French, German and various languages of the Turkish family. We argue that this adaptation does not support the perceptual view, which maintains that loanword adaptations are due to L1 “deafness” to L2 foreign material and the fact that L1 borrowers reconstruct what is natural to their ear when they are in L1 mode.

Similarities and differences in the perception of formant frequency transitions and center-of-gravity glides

In Lublinskaja (1996) the spectral ‘‘center-of-gravity’’ (COG) effect was observed in the identification of synthetic vowels in which the amplitude ratio of F2 and F3 was gradually changed in time, formant frequencies being constant. The stimuli with such virtual formant transitions (VT) were perceived as diphthong-like Russian vowels when F3−F2&amp;lt;4.5 Bark. Within this F3−F2 frequency separation, the stimuli with the dynamic perceptual formant (F2*) were identified as their equivalent vowels with real F2 and F3 transitions (FT). But the perception of both types of transitions was not identical. Generally, the VT signals were less salient than FT. Within smaller F2−F3 separations (F2*=1.68 Bark), the VT signals were first identified with 100%-accuracy as diphthong-like vowels and with larger F2* the glide transitions were gradually less perceptible. With separations F3−F2&amp;gt;4.5 Bark the VT stimuli were identified as stationary vowels, and listeners’ decisions were based on terminal values of F2*. Reduced salience of VT glides relative to FM tones was also reported for non-speech signals [Dawson and Feth, ARLO 5, 95–99 (2004)]. The talk will discuss possible mechanisms for why VT signals are less perceptually prominent than FT sounds.

Within- and across-language spectral and temporal variability of vowels in different phonetic and prosodic contexts: Russian and Japanese

Contextual variability of vowels in three languages with large vowel inventories was examined previously. Here, variability of vowels in two languages with small inventories (Russian, Japanese) was explored. Vowels were produced by three female speakers of each language in four contexts: (Vba) disyllables and in 3-syllable nonsense words (gaC1VC2a) embedded within carrier sentences; contexts included bilabial stops (bVp) in normal rate sentences and alveolar stops (dVt) in both normal and rapid rate sentences. Dependent variables were syllable durations and formant frequencies at syllable midpoint. Results showed very little variation across consonant and rate conditions in formants for /i/ in both languages. Japanese short /u, o, a/ showed fronting (F2 increases) in alveolar context relative to labial context (1.3-2.0 Barks), which was more pronounced in rapid sentences. Fronting of Japanese long vowels was less pronounced (0.3 to 0.9 Barks). Japanese long/short vowel ratios varied with speaking style (syllables versus sentences) and speaking rate. All Russian vowels except /i/ were fronted in alveolar vs labial context (1.1-3.1 Barks) but showed little change in either spectrum or duration with speaking rate. Comparisons of these patterns of variability with American English, French and German vowel results will be discussed.

The Process of the Fall of the Reduced Vowels in Old Russian in the Light of the Piotrovskii Law

The Process of the Fall of the Reduced Vowels in Old Russian in the Light of the Piotrovskii Law

Spectral Dynamics and Classification of Russian Vowels

Spectral Dynamics and Classification of Russian Vowels

Experiment on investigating the voice cords functioning by the speech signal

A practical application of linear prediction methods for calculating the pulse function that models the functioning of vocal cords is described. Some characteristics of the pulses of this function enable one to draw some conclusions about the speaker’s individual features (and, possibly, about the quality of sound). The first part of the paper is devoted to the theoretical background of the described method. The second part presents a detailed algorithm of the program realization of the method in the MATLAB 5.2 environment and analyzes the results of the experiment made on Russian vowels.

Dynamic center-of-gravity effects in consonant–vowel transitions

Lublinskaja has reported that dynamic changes in the spectral center-of-gravity (COG) of one formant in selected Russian vowels led to changes in the identification of the vowel. Movement of the COG was effected by simultaneous amplitude modulation of two formants placed at the end points of the desired frequency transition [Iyer et al., J. Acoust. Soc. Am. 109, 2294 (2001)]. The present study explores whether such COG effects will extend into the processing of consonant–vowel transitions (CVs). A [dɑ]–[gɑ] continuum was generated using a Klatt synthesizer with a 50-ms CV transition and a 200-ms vowel. The onset of F3 varied from 2018–2818 Hz in 80-Hz step. The F3 of the vowel remained constant at 2527 Hz. Three different stimulus sets differed in terms of whether the F3 transition was a: (1) formant, (2) FM tone, or (3) virtual frequency (VF) glide. In the VF stimuli, the frequencies of two tones matched the onset or offset frequencies of the F3 transition and modulation of their amplitudes produced a gliding spectral COG. Preliminary results indicate that listener performance is not strongly affected by changing the means by which spectral changes to F3 are made. [Research supported by a grant from OSU’s College of Social and Behavioral Sciences to L. Feth and an INRS award from NIH to R. Fox.]