LINGUIST List 23.4556

Wed Oct 31 2012

Review: Computational Linguistics; Phonetics: Ladefoged & Ferrari Disner (2012)

Editor for this issue: Rajiv Rao <rajivlinguistlist.org>



Date: 31-Oct-2012
From: Seetha Jayaraman <seetha.jaygmail.com>
Subject: Vowels and Consonants
E-mail this message to a friend

Discuss this message

Announced at http://linguistlist.org/issues/23/23-2477.html

AUTHOR: Peter Ladefoged and Sandra Ferrari DisnerTITLE: Vowels and ConsonantsEDITION: ThirdPUBLISHER: Wiley-BlackwellYEAR: 2012

Seetha Jayaraman, Dhofar University, Sultanate of Oman

SUMMARY

This book, written by Peter Ladefoged and revised by Sandra Ferrari Disner,contains sixteen chapters on topics ranging from the basics of speech sounds toan advanced description of acoustic features and the role of computers in studyingacoustic components of speech. The chapters cover perspectives on speechproduction and perception and give an overview of phonatory and articulatoryprocesses involved in the production of different categories of speech sounds,viz., vowels and consonants. The last three chapters deal with articulatorydifferences found in different languages around the world.

The volume provides an exhaustive list of illustrations of sounds discussed ineach chapter and audio-recordings, photographs and videos of vocal tractconfigurations are made available on the websitewww.linguistics.ucla.edu./faciliti/sales/software.htm. A table lists the audio-recordings supporting the volume that are available on the website.

A chapter-wise summary follows:

Chapter 1, "Sounds and Languages", begins with the definition of 'sound' and thedistinction between 'sound' and 'language'. It discusses how languages evolve anddisappear constantly with changes in the socioeconomic conditions of people andtheir cultural practices. It reflects on the importance of individual sounds, differentaspects of language and speech, and the role they play in our life. The chapteralso describes speech sounds and sound symbols (i.e. International PhoneticAlphabet) vis à vis their orthographic representations, with an introduction to thebasic components of speech, viz., pitch and loudness and their representations ina waveform.

Chapter 2, "Pitch and Loudness", discusses 'tones' in terms of pitch and meaningchange associated with pitch, drawing upon examples from tone languages likeChinese (Mandarin) and Cantonese. The fundamental concepts in understandingpitch levels, pitch curves and intonation, with reference to the speaker, areexplained. The last section of this chapter outlines the importance of vocal folds,their position in sound production and the influence of vocal fold vibration onloudness, in general, and on English intonation, in particular.

The next three chapters (3,4 and 5) present a description of vowel features, thevowel chart, the vowel space, and acoustic characteristics that help identifyvowels in spectrograms with respect to the structure of the first three formants.

Chapter 3, "Vowel Contrasts", compares vowels across languages like Spanish,Hawaiian, Swahili and Japanese in order to bring out differences in their usage:some examples are 'masa' (dough) and 'mesa' (table) in Spanish; 'kaka' (to rinse)and 'keka' (turnstone) in Hawaiian; 'pata' (hinge) and 'peta' (bend) in Swahili; and'ma' (interval) and 'me' (eye) in Japanese. This chapter also highlights thedifferences between General American English and Standard British English intheir use of vowels. General American English consists of only 14 or 15 vowels,while British English consists of as many as 20 vowel sounds.

Chapter 4, " The Sounds of Vowels", gives an account of both the acousticcharacteristics of vowel quality and formant patterns in spectrograms as evidencefor vowels. There is a detailed explanation of the interplay between the first twoformant values and the vowel space. When the pitch changes associated withvowel changes are plotted in a graph with F1 and F2 (frequencies, as we hearthem in different languages), the resultant figure is a triangle. Given that theauditory space for the three possible vowels /i/, /a/, and /u,/ the vowel space in thegraph shows a triangular shape. With languages having 5 to 7 vowels, it ispossible to have an equally symmetrical triangular shape when we plot F1 vs. F2;this same shape for any language provides evidence of a relationship betweenvowel quality and formant frequencies.

Chapter 5, "Charting Vowels", continues the discussion on formant analysis andcharting of vowels through the first two formants, comparing the five vowels ofSpanish with those occurring in different accents in English. The relative vowelspace plotted for the Spanish vowels /i, e, a, o, u/ is compared with that ofEnglish. There is a tendency to replace diphthongs with their correspondingmonophthongs in some North American accents. With the exception of the vowelin words like 'bird', the third formant is not significant in the description of vowelsin General American English.

The next chapter, "The Sounds of Consonants", provides an introduction toconsonants and suggests that there is no significant difference in consonantarticulation between British and American varieties of English. The phoneticsymbols used and the articulatory and acoustic features of consonants aredescribed. This chapter provides background information on different classes ofconsonants, viz., stops, approximants, nasals, fricatives and affricates.Interpretation of spectrograms with respect to both voiceless and voicedconsonants is explained as well. The first three formant frequency values, theirlevels, formant transitions for stops, nasals and approximants, and additionalspectral cues which help in the identification of individual consonants areillustrated with examples from General American English and BBC English.

Chapter 7, "Acoustic Components of Speech", analyzes formant frequency,amplitude and pitch, combining and varying their auditory correlates of voicing andvoicelessness. Speech synthesis is also discussed, as well as and therelationship between acoustic variables in the waveform, which are illustrated forthe English word 'bird'.

Chapter 8, "Talking Computers", continues with the topic of synthesizing speechsounds, with phonetic transcription being the focus of the last part of this chapter.Two approaches to speech synthesis are suggested: parametric synthesis, wherea computer calculates acoustic parameters like formant frequencies from thewaveform or joins sound segments to make new sentences; the concatenativeapproach, in which large sections of speech are stored and subsequently joinedtogether. The problem with the first approach is that we do not know the rules ofjoining one sound to another. The second approach is useful in synthesizingrecordings of telephone numbers and reproducing them for providing pre-recordedinformation. The computer uses a mathematical technique called Linear PredictionCoefficient (LPC) analysis, which uses LPCs, or a set of numbers that representeverything about voice quality except its fundamental frequency or pitch. Adetailed account of LPC analysis is also given in Ladefoged (1996). Anothersystem called Pitch Synchronous Overlap Add (PSoLA) is also employed either bylowering or raising the pitch of the original recording or by recording the variation induration. The last section of the chapter deals with studying segmental errorswhen using Text To Speech (TTS) systems in intonation. Spelling out allabbreviations and numbers using IPA symbols is a prerequisite in TTS.

Chapter 9, "Listening Computers", contains an account of the way sounds arerecognized and displayed on a computer. The chapter illustrates the spectralrepresentation of the first three formants in the word 'August'. Identifying individualsounds with spectral cues is another dimension viewed in this chapter. The authoracknowledges Fred Jelinek's contribution to speech recognition and lists out thestages involved in the speech recognition system. He also considers the term'cepstral coefficient', which refers to measures of spectral slices stored as anumber and reflects the rise and fall in the amplitude of F1, F2 and F3 in aspectrum or spectral curves. Computers also use the Hidden Markov Model(HMM), which is a representation of a sequence of speech events.

Chapter 10, "How We Listen to Speech", deals with different ways of listening forphonetically confusable sounds that impede intelligibility. A confusion matrix forsyllables with different initial consonants and noise levels is shown on a table. Thepremise of the table is the way these sounds are heard by a set of listeners. Theconfusion matrices tell us the level of confusion and the degree of similaritybetween the sounds using the syllables 'pa', 'ta', 'ka', and so on. The higher thenumber of correctly heard syllables, the less confusion there is. Perceptualdifferences are calculated using 16 sets of syllables. This chapter also reports theresults of an experiment conducted with the words 'bad' and 'bat' (voicing contrast)to study variation in perception. This is the only chapter that provides sources forfurther reading on the topics discussed.

Chapter 11, "Making English Consonants", deals with the physiology of the vocalapparatus and the articulatory terms associated with the description of place andmanner of articulation of consonants in general. The table of IPA symbols ofEnglish consonants is presented with a brief description of each class ofconsonants.

Chapter 12, "Making English Vowels", describes the anatomy and physiology ofvocal organs and the muscles controlling the movements of the tongue in theproduction of vowels. There is an interesting account of Melville Bell's symbols, asgiven in his Visible Speech (1867), representing vowels in English. The positionand shape of the tongue and palate in the production of vowels relating to thevowel diagram are analyzed in detail.

Chapter 13, "Actions of the Larynx", talks about the important role played by thelarynx, pharynx, vocal folds, and cartilage and the changes they bring to thequality of sounds (viz., voiced and voiceless sounds). Voicing and aspiration aretwo important features in the production of stop consonants. An important featureamong these is aspiration and Voice Onset Time (VOT), which vary amongstlanguages. The interval between the release of a stop and the beginning of thefollowing vowel is called Voice Onset Time (VOT). In English VOT is 50-60milliseconds (ms) for /k/ and slightly less for /t/ and /p/, while in Spanish the VOTfor /k/ is about 20 ms and even less for /p/. It is interesting to note that Germaniclanguages like English, German and Danish have comparatively longer VOTs. InRomance languages like French and Spanish, there is no VOT of voiceless stops,while English and other Germanic languages have voiced stops, which contrastwith voiceless stops. In terms of vocal fold vibration, glottal stop consonants like/h/ are found to be replaced by /k/ or /p/ in some dialects of British English, as wellas in Hawaiian. Examples from Hindi also show the occurrence of four breathyvoiced stop consonants, while Gujarati has breathy voiced vowels. The effect ofcreaky voice and breathy voice on Zapotec vowels is discussed briefly. Otherclasses of sounds discussed are 'ejectives', common in a few American Indianand a few African languages, and 'implosives', which are produced with air suckedin and found in some languages spoken in Nigeria (e.g. Owerri lagbo). Themechanism involved in producing implosives is illustrated through differences inairflow and air stream in the larynx and the vocal tract.

Chapter 14, "Consonants Around the World", is a summary of consonants inlanguages. A general survey shows that there are about 7,000 languages in theworld and over half of them are spoken by fewer than 10,000 people. In all, thereare about 600 consonants. The chapter lists the 10 most widely spoken languageswhich have 100 consonants (of which, 22 occur in English). A few languages likeEwe, spoken in Ghana, use two unique bilabial fricatives. Subtle differences whichexist in the production of /t/ in Wabuy (a language spoken in Australia) palatals inHungarian, stops and six nasals in Malayalam, voiceless stops in Aleut, andbilabial and alveolar trills in Kele and Titan, respectively, are detailed. Likewise,F1-F2 transitions (palatals) in palatograms and linguagrams of the retroflex /ţ/,Polish sibilants and four sibilants of Toda (and their corresponding IPA symbols),are also discussed exhaustively. Laterals in Melpa are noted for their manner oftheir articulation, as they are complex in symbols, viz., voiced alveolar /l/ andvoiceless velar /ł/ (dark /l/, represented by a small uppercase L in IPA). In Zulu,laterals occur as voiced and voiceless consonants and clicks occur contrastively.Nama, a language spoken in Namibia, has 20 clicks, each represented by an IPAsymbol and with different meanings.

Chapter 15, "Vowels Around the World", demonstrates the relation between vowelspace and the graphic display of F1-F2. Contrasts are made between languageslike Hawaiian, which has 5 vowels and only 8 consonants, and those such as Zulu,which has 5 vowels and 44 consonants. Every language is said to use at least 3distinct vowels, viz., /i , a, u/ or /i, a, o/. About 20% of the world's languages have5 contrasting vowels. An interesting fact is that most languages with 5 vowelsfollow the same order of the Latin alphabet (a, e, i, o, u). Californian English has 15vowels and BBC English has 20 vowels (12 long, 10 short and 6 diphthongs) withvarying tongue roots. Lip rounding also plays an important part in the articulation ofvowels. French has rounded vowels like /y/, as in 'lu' (a front, high, roundedvowel). The other rounded vowels which occur in French are /œ/ as in 'leur' (their),/ø/ as in 'le' (the), /o/ as in 'lot' (prize), /ɔ/ аs in 'lors' and /ɑ/ as in 'las' (tired).Swedish, Danish, Norwegian and German also have rounded/unrounded vowelcontrasts. Nasal vowels versus nasalized vowels are observed in English andFrench, respectively, as in the vowels in the words 'lin' (flax), 'lundi' (Monday),'lent' (slow) and 'long' (long). Phonetic differences in vowels are observed withdistinctions in voice quality, as in !Xóō vowels (a Bushman language spoken in theKalahari desert) or tense-voiced vowels in Mpi (spoken in Northern Thailand).

The last chapter in this volume, "Putting Vowels and Consonants Together",summarizes vowels and consonants, puts them together as 'utterances', and talksabout the speech continuum in terms of duration and intelligibility. It is a commonobservation that slips of the tongue, which interchange the sounds of syllables,occur in speech. The other aspects discussed are writing systems and sounds,tones and languages like Chinese (Mandarin) and Cantonese. The role of IPA inrepresenting /r/ and its variants in languages other than English, contrastingsounds, and so on, are emphasized. In all, 106 distinct symbols for segments (78consonants and 28 vowels), excluding sounds like ejectives and diacritics, arerepresented in the IPA chart provided. Sounds are also transcribed using symbolslike )( (not an IPA symbol) for 'hiss' or 'sing'. The totality of features required todescribe a language at a glance is shown in a single table (Table 16.2 on page196).

EVALUATION

The book is an excellent introduction to the basics of speech sounds. The numberof books available on phonetics is innumerable, but "Vowels and Consonant" isundoubtedly one of the best books on the basics. It is a good example of howcomplex topics like acoustic phonetics, speech synthesis, speech recognition, thephysiology of speech production and sound-spelling correlation can be simplified tobe accessible for beginners in phonetic studies. It requires and assumes no priorknowledge, either of phonetics or the process of speech production, on the part ofthe reader. Each chapter is introductory in nature and technical terminology hasbeen used sparingly while explaining the basics of both articulatory and acousticphonetics. The topics cover a wide range, from traditional definitions of phoneticterms and an IPA chart, to the latest trends in TTS systems used in speechtechnology. The last three chapters are dense and rich in content, and consonantand vowel sounds across different languages of the world (the most widelyspoken) have been discussed extensively, clearly and concisely.

Chapter 6 is especially effective because it equips the reader with all the details ofconsonant features with remarkable clarity and precision. Chapters 14 and 15 ofthe volume also merit special mention due to their coverage of examples from allthe distinctive sounds of a few lesser known, yet widely spoken languages. Thedetail in these two chapters aptly justifies the title of the book.

The volume is a valuable contribution for researchers and scholars working onconsonants and vowels across different languages. It serves as a goodintroductory textbook for a course on phonetics. The highlight of the third edition of"Vowels and Consonants" is the demos of some Text-to Speech Systems such asvideos of vibrating vocal cords, audio recordings of articulations of vowels andillustrations of IPA symbols. As stated in the Preface to the Third Edition, "The CDthat had accompanied the previous edition has been replaced with a more readilyaccessible web-based collection of language files" (p. xv). The volume serves as aready reference for advanced users of phonetics, as well as professionals andresearch scholars of language and speech. The book is of interest to teachers andwould help to develop readers' perception of speech production and theircompetence in spoken English. It is a 'must have' book that adds richness andknowledge to individuals and libraries.

REFERENCES

Ladefoged, P. (1996). A Course in Phonetics, (2nd Ed.). Chicago, ChicagoUniversity of Press.

ABOUT THE REVIEWER

Dr. Seetha Jayaraman is a Lecturer at Dhofar University, Sultanate ofOman, where she teaches English language to undergraduates. Herresearch interests include sociolinguistics, musicology, comparativelinguistics, and phonetics.

Page Updated: 31-Oct-2012