Vowel Classification Research Articles

The IPA alphabet: a case for reform

It is surely incontestable that the IPA alphabet, like the system of vowel and consonant classification on which it is based, contains obvious anomalies; some or others of which have been discussed in recent proposals for reform (e.g. Breckwoldt 1971, McClure 1972, Hammarström 1973). Fox (1975), in an interesting attempt to widen the scope of the discussion, has argued that the anomalies are less serious than they appear, on the grounds that (a) they do not interfere with the practical usefulness of the alphabet, and (b) a more positive and more fundamental point—they reflect in some cases what appear to be linguistic universals, such as the potential use of retroflex consonants as a distinctive category and the widespread preference for voiced over voiceless nasals.

Quantal configurations for vowels

Examination of the vocal-tract configurations that are available for the production of vowels, and of the properties of the sounds produced by these configurations, suggests that certain configurations are preferred in the sense that they give rise to sound outputs that are distinctive and whose properties are relatively insensitive to perturbations in the articulatory representation. It is shown that anterior-posterior movement of the tongue body leads to two such preferred configurations independent of tongue height-configurations corresponding to the traditional phonetic feature [± back]. Variation in tongue height is postulated to produce three distinctive positions achieved through movements of the tongue and the mandible: tongue maximally raised toward the hard or soft palate, tongue lowered but with some palate contact maintained with the edges of the tongue, and severed contact of the tongue with the palate. Shifts from one of these positions to another produces well-defined contrasts in the acoustic output. Quantal adjustments leading to distinctive acoustic contrasts can also be made in the lips and tongue root, Implications of this point of view in the formulation of systems for the classification of vowels are discussed. [Supported in part by an NIH Grant.]

A prosodically guided speech understanding strategy

Our strategy for computer understanding of speech uses prosodic features to break up continuous speech into sentences and phrases and locate stressed syllables in those phrases. The most reliable phonetic data are obtained by performing a distinguishing features analysis within the stressed syllables and by locating sibilants and other robust information in unstressed syllables. The numbers and locations of syntactic boundaries and stressed syllables are used to select likely syntactic and semantic structures, within which words are hypothesized to correspond to the partial distinguishing features matrices obtained from the segmental analyses. Portions of this strategy have been implemented and tested with hundreds of seconds of speech, involving fifteen talkers. A program for detecting syntactic boundaries from fall-rise patterns in fundamental frequency contours correctly detected ever 90 percent of all predicted boundaries. An algorithm for locating stressed syllables (from fundamental frequency contours and high-energy syllabic nuclei) correctly located the nuclei of over 85 percent of all those syllables perceived as stressed by a panel of listeners. A study of segmental analysis results obtained by several other research groups showed that phonetic recognition clearly is most successful in the stressed syllables. Procedures for classification of stressed vowels, location and classification of sibilants, and location of stops, nasals, and [r]-like sounds have been implemented. Prosodic aids to parsing and semantic analysis are being investigated.

Parsing and word matching in the Lincoln Laboratory, speech recognition system

The Lincoln system is designed to recognize sentences conforming to a context-free grammar. Parsing is guided by a heuristic evaluation function which combines individual word scores into parse path scores. The word scores measure the degree of correspondence between phonemic dictionary spellings and the results of an acoustic-phonetic analysis of the input sentence. Word scoring is based on two computer-generated scoring matrices derived from confusion statistics gathered from 113 sentences. The synchronization problem resulting from missing and spurious segments is simplified by first aligning the vowels (sometimes in more than one combination). The effect of phonological rules is handled by insetting optional phonemes in the dictionary and flagging others as possibly missing. Other rules, which are specific to Lincoln's front-end analysis, predict spurious segments and the effect of neighboring semivowels on vowel classification. [This work was sponsored by the Advanced Research Projects Agency of the Department of Defense.]

Some Results on Segmentation and Segment Classification in Continuous Speech

A system for acoustic analysis of continuous speech is being developed as part of Lincoln Laboratory's effort in speech understanding. The acoustic system presently consists of two stages: the first produces a preliminary segmentation of the speech, as well as a gross classification of these segments into several acoustic categories; the second produces a finer classification intended to meet the needs of subsequent linguistic analysis. The current design and performance of the acoustic analysis system is outlined. This description is intended primarily as an interim report, since the system is being actively extended and refined. The first-stage segmentation and segment classification yields the following categories: vowel-like sound; volume dip within vowel-like sound; fricative-like sound; silence or voice bar; and burst. These categories are produced by a decision tree based upon energy measurements in selected frequency bands, derivatives, and ratios of these measurements, a voicing detector, and a few editing rules. Statistics will be presented describing the results of this analysis on a 31-sentence test corpus developed for an ARPA Segmentation Workshop held at Carnegie-Mellon University in July 1973. Second-stage classification includes a fricative identifier, which employs spectral measurements to group the fricative segments into phoneme-like categories. Results of this analysis for the test corpus will be presented. Discussion of additional work, including stop and vowel classification, will be given as time permits. [This work was sponsored by the Advanced Research Project Agency of the Department of Defense.]

Syllable Recognition Using an Adaptive Pattern Recognition Technique

Sonagrams of CVC (C = consonant, V = vowel) syllables spoken by a single male speaker, and VC, CV, CVC syllables spoken by two male speakers were converted into 140-dimensional and 144-dimensional quantized patterns respectively. It has been shown that when the components are optimally weighted before being summed, the weighted sum can serve as the basis of recognition. An index of correct recognition has been defined, and a rule for arriving at the correct weights of the individual components has been stated. The sample on which the work has been done is small. Nevertheless, it is of some significance that a procedure of this kind may lead to the identification of vowel classes, syllables and even speakers. The use of adaptive pattern recognition technique is suggested for the recognition of the weighted patterns and an automatic weight determining device is proposed to optimize the weights during the learning phase of the machine.

An Electromyographic Investigation of the Feature of Tension in Some American English Vowels

Both the traditional vowel triangle and more recent schemes of vowel classification present us with a picture in which the members of at least three pairs of vowels, /i-I/, /e-ɛ/, and u-U/, are said to be differentiated from each other by a feature of tenseness. It has been suggested that differences in tension between the members of a tense-lax pair of vowels are manifested in the extrinsic muscles of the tongue. In the experiment reported here, electromyographic potentials for the genioglossus muscle were investigated to determine the extent, if any, of tension differences between the members of the pairs of vowels listed above. It was found that the posited differences in tension do exist, although the findings cast some doubt on traditionally proposed height relationships for the front series of vowels, /i-I-e-ɛ/. Electromyographic potentials of muscles associated with jaw opening and tongue backing were investigated to see if the genioglossus data could be reconciled with the usual height ordering of the front vowels. Results indicate that there is reason to believe the height ordering of these vowels generally presented in the literature is correct.

On the Classification of Russian Fricatives in CV Syllables for Different Speakers

The sound spectra of the voiceless hard and soft Russian fricatives have been studied in various vowel contexts and with a population of five different speakers (three men, two women). The acoustic characteristics of the fricatives are described in terms of distributions in the FC, FL-plot, where FC is the frequency of the centroid of a short-time spectral section of the voiceless part of a fricative, and where FL is the frequency of the very beginning of the second formant transition from a fricative to a vowel. It is shown that the proper classification of the fricatives requires the self-normalization procedure, because of considerable differences which occur between the same fricatives spoken by different speakers. Results obtained from the study confirm the efficiency of the statistical method of self-normalization for the classification of the fricatives. The method was suggested earlier by the author for the classification of vowels.

Classification of Russian Vowels Spoken by Different Speakers

The problem of speaker normalization is investigated for classifying the Russian vowels. The known methods of normalization of formants F1 and F2 which make their normalized values F1N and F2N invariant to compression or expansion and to parallel shift are mentioned. A new statistical method of normalization is then suggested. All of these methods are compared using an index of normalization quality (η). The result of the comparison shows that the normalization suggested in this letter has the largest index, not only on the average but also for each individual vowel pair. The procedure of classification of the Russian vowels using a self-normalized formant plot is shown at the end of the letter.

Effect of intertrial interval and verbal reinforcement on choice of front or back tongue position.

Intertrial intervals of 2 sec. as compared with 4 sec. significantly increased choice of front-articulated vowels as compared with back-articulated vowels in 4 reinforced and 4 non-reinforced groups of 5 Ss each. Ss who were native English speakers showed a significant preference for front over back-articulated vowels. The hypothesis that shorter intertrial intervals would increase conditioning was not supported since no reinforcement effects were found. Discrimination between the two response classes of front and back-articulated vowels was demonstrated in the 4 2-sec. intertrial interval groups as compared with the 4 4-sec. intertrial interval groups. The preference for front over back-articulated vowels and increased preference for the front over back-articulated vowels due to shorter intertrial intervals was discussed in terms of stimulus choice due to prior learning of English.

The Phonetics of Russian

Preface Abbreviations Part I. Phonetic Theory and Transcription: 1. Introduction 2. The organs of speech 3. Breath and voice 4. Vowels and consonants 5. How to learn vowels 6. The classification of vowels 7. How to describe consonants 8. The principles of transcription, phonemes 9. Digraphs Part II. the Phonetics of Russian: 10. Types of Russian pronunciation 11. General notes on the Russian vowels and consonants 12. The Russian vowels in detail 13. The Russian semi-vowel j 14. The Russian diphthongs 15. Labialization and velarization 16. Palatalized consonants and their phonetic representation 17. The Russian plosive consonants 18. The Russian fricative consonants 19. The Russian affricative consonants 20. The Russian nasal consonants 21. The Russian lateral consonants 22. The Russian rolled consonants 23. Similitude and assimilation 24. Accent, stress and length 25. Intonation 26. Passages for practice Appendices Indices.

Classification of self-normalized vowels

The classical vowel measurements of Peterson and Barney in 1952 have been reexamined on an individual basis to determine the minimum dimensionality required to classify a single talker's vowels, and the minimum number of vowels required to specify the dimensions. It was found that two formants are adequate for this purpose, and that no more than three vowels are required to scale the frequencies of two-formant space. Within that self-normalized space, all vowels have standard locations which are similar for all talkers.

Study of Vowel Classification

It is generally accepted that the voiced portions of speech may be idealized into time variations of the first three formant frequencies and the fundamental frequency. The influence of these parameters upon the classification of short steady-state vowel segments was determined. The mechanism by which a listener quantizes the voiced portions of speech into units subjectively equal to the steady-state vowel segments is presented. The application of this method of vowel classification to an automatic speech recognition process is discussed.

A Phonologic and Acoustic Classification of Polish Vowels

A Phonologic and Acoustic Classification of Polish Vowels

Unsolved Problems in Spanish and Spanish-American Linguistics

Despite the enormous progress that has been achieved in linguistic studies, and especially in Romance linguistics, during the last half century, we can certainly say that thei most striking feature of research in Hispanic philology is the fact that so little work has been done, by and large, in a field that is potentially so rich. * There are notable exceptions, of course, as in the case of certain areas and topics; but as yet the field of Hispanic linguistics lacks the basic works, and even the research tools and monographs, that will make possible some day the great dictionaries and histories that the Hispanic languages so richly deserve. There are many different approaches to philological study, and many different aims. Let me say from the outset that I have little interest in the controversies of the several linguistic schools. We can be sure that the collaboration of all is necessary if we aspire to obtain a complete picture of the linguistic history of the Hispanic peoples; the successful investigator, in our field as well as in others, must work from a broad basis and, like the anthropologist, for example, use not only the tools necessary for minute analysis and historical reconstruction, on the one hand, but also explain and evaluate in terms of meaning and function, on the other. In reviewing the various fields represented in my report, I shall take up first the traditional disciplines, such as phonetics, phonology, morphology, and syntax, and then discuss briefly our needs in some of the newer fields, like dialectology, linguistic geography, and stylistics. I shall conclude by discussing some of the more general needs in the field, such as bibliogaphies, surveys, seminars, and the training of investigators. Sta ting, then, with the field of phonet s, it can be said that in no branch of linguistic studies has such progress been at ained as in the study of speech sounds. The key to the historical interpretation of the development of language has been found principally in this discipline, a field in which a great number of discoveries have been made and new methods devised. In the Hispanic field, however, despite the cognized importance of phonetics in linguistic studies, and despite the exemplary studies we have concerning Peninsular Spanish, only a small part of Hispanic America has been studied scientifically. And even where some work has been done, a change in method is at times necessary. One of our most urgent needs is a series of manuals on the pronunciation of the principal linguistic regions of America comparable to that of Professor Navarro for Peninsular Spanish. For regions as important as Mexico or Colombia we still lack an accurate description and classification of vowels and consonants. In the case of vowels we need to know more about their quality and quantity, weakening of unstressed vowels, tendencies toward diphthongization, labialization, or nasalization. The treatment of groups of vowels is especially important. As I suggest below, the study of the versification of folk poetry would be of assistance in dealing with this question. * A paper read before the Spanish Language and Mediaeval Literature Section of the Modern Language Association at its Annual Meeting in New York City, December 29, 1950. 233

The Relation to Hittite Verbs in -a(i)- to the Thematic Vowel Class

For the Hittite verbs with stems in -a(i)-, e.g. hatr-a-mi 'I write' 3 sg. hatraizzi, or handami, handaizzi 'establish', what is perhaps an acceptable explanation was found by G*tze, Madduwattas 97 (Hetitische Texte in Umschrift, herausgeg. v. F. Sommer, Heft 3) and Sturtevant, LANG. 5.9 f. Starting from the fact established by the former, that the ai-forms occur only where the IE thematic vowel was e, and that those forms which had the o-grade always have -a-, the latter concluded that the a(i)-class represented the IE denominatives in -d-jo-. In hatranzi < *hatrayanti IE *-a-jo-nti, for instance, the y dropped between two vowels of the same quality, while the ending of hatraizzi comes from IE *-da-e-ti. While none of the objections which might be urged against this theory are in themselves insurmountable, another explanation which avoids these difficulties may appear more convincing, and consequently, with due reservations, I suggest below an alternative which seems to me to be beset with fewer difficulties.

Singing, Speaking, and Stammering

IN NATURE, vol. xxvii. p. 532, in the report of Dr. Stone's lecture on “Singing, Speaking, and Stammering,” there appears a Classification of Vowels, which is described as an abstract of Mr. Melville Bell's scheme. I should like, however, to point out that the system which Mr. Bell has advocated for the last fifteen years is hardly represented in the Classification referred to. On turning to p. 63 of Mr. Bell's “Sounds and their Relations,” which is a new exposition of “Visible Speech,” it will be seen that the vowels Ir and Ah are not described as labio-lingual, and, that the threefold arrangement of the vowels as lingual, labio-lingual, and labial is abandoned as incorrect. The lecturer does not appear to have mentioned the phonetic researches of Mr. A. J. Ellis and Mr. Henry Sweet. In many important points, however, they supplement the system of Mr. Bell, and their works cannot be overlooked in the scientific study either of etymology or pronunciation. The student of language can hardly do better than begin with Mr. Ellis's “Speech in Song,” and Mr. Sweet's “Handbook of Phonetics”.

Singing, Speaking, and Stammering

THE classification of vowels to which Mr. Lecky refers is taken from Mr. Melville Bell's “Principles of Elocution,” which I obtained with much difficulty from a publisher in Salem, Massachusetts. It is dated 1878, and may, I suppose, be held to represent the author's system at that date. I am well acquainted with the other works named by Mr. Lecky.