Operatic IPA and the Visual Thesaurus

« previous post | next post »

In my new capacity as executive producer for the Visual Thesaurus (a job title Mark Liberman had some fun with), I'm responsible for editing the content of the website's online magazine and also for creating some of it. I've just launched a new column called "Word Routes," which I'll be posting a couple of times a week. (The column is freely available, but to leave comments or to take advantage of all the other Visual Thesaurus goodness you need to sign up for a subscription.) The first installment of Word Routes is on the word procrastination, which is also the subject of an article I wrote for today's issue of Slate. But rather than just toot my own horn, I wanted to draw attention to some great work that was recently done for the Visual Thesaurus, harnessing the unexpected prowess of opera singers to read the International Phonetic Alphabet.

As profiled in this announcement, the Visual Thesaurus enlisted the help of some vocally talented freelancers to make recordings for the 150,000 words and phrases in VT's lexical database (in turn drawn from WordNet). To my knowledge, no phonetic recording project of this scope has ever been attempted for an English-language online reference. As I mention in the announcement, we're talking about 55 hours of recorded speech, all done in tiny chunks averaging about 1.3 seconds. It turned out that the ideal candidates for this sort of freelance project were performers with a background in opera — for their vocal stamina, attention to the details of enunciation, and most of all, knowledge of IPA. As it happens, just about every opera singer and coach needs to be familiarized with IPA in order to sing in non-native languages, using phonetically rendered libretti. There are occasional exceptions like Pavarotti, who only sang in his native Italian (and, it was whispered, couldn't read music either). But everyone else picks up IPA during their training, whether in a formal conservatory setting or less systematically in private study with coaches.

For the opera singer in training, there are a number of resources available for IPA study. Nico Castel publishes a libretti series covering most of the major operas, employing the "Leyerle Three-line Phonetic-Translation System" (IPA representation on the first line, original language on the second, and word-for-word English translation on the third). For operas and song literature not covered by Castel, there's IPA Source, a vast online library of transcriptions. Or the enterprising young singer might try to do his or her own IPA transcription of a particular piece, using, for instance, German dictionaries from Duden that provide IPA. For relatively uncommon languages like Czech, few phonetic resources are available, so much of the vocal preparation has to be self-directed by necessity.

IPA only gets you so far, however, since transcribed libretti might not provide a terribly "close" phonetic transcription. Vowel length, for instance, is not always indicated in IPA renderings, and may vary from language to language. But armed with IPA, singers can work with coaches (who often specialize in a particular language) to tackle aspects of vocal performance not encompassed by the transcription. Needless to say, after that sort of intensive training, pronouncing IPA representations of citational forms in one's own language (as the Visual Thesaurus team did) is comparatively easy work!

(Thanks to Sara Antunovich for her operatic insights.)

[Update: Mark Liberman covered some of this terrain in two posts way back in July 2004.]


  1. Bobbie said,

    May 14, 2008 @ 3:09 pm

    When I was in college, I studied phonetics and was in the University Chorus. Our choir director instructed us on the "proper" way to pronounce/sing many words — with "open" vowels, open mouths, and with full breaths from our diaphragms. Since very few of the chorus members were familiar with IPA symbols, he spent a lot of time demonstrating how to sing each word. I incorporated all the coaching in my term paper for my phonetics class that semester, which compared "everyday" pronunciation with operatic pronunciation.

  2. john riemann soong said,

    May 14, 2008 @ 4:01 pm

    Since IPA provides a means of describin slack voice, harsh voice, etc., does "operatic IPA" incorporate the different opera voice registers?

  3. Ryan Stickney said,

    May 14, 2008 @ 5:31 pm

    @john riemann soong: No. As an undergraduate double-major in classical voice and linguistics, I can tell you that "Operatic IPA" is relatively impoverished, failing to make many vowel distinctions even in commonly-sung languages like the staple Italian, German, and French. This may be because the transcribers figure the vowels will be distorted anyway due to the different requirements of vocal production in various registers (and this is often true; vowels are frequently modified due to perceived or actual advantages in hitting notes). It certainly does not incorporate different registers, which are not indicated by breathy, creaky, or otherwise modified vocal tone marks anyway. Opera singers' use of IPA (in my experience so far) tends to be more toward the sketchy side. Clearly it's useful for words you don't know in your native language, though. I bet that was a fun job.

  4. Pamela Hay said,

    May 15, 2008 @ 9:23 am

    Perhaps IPA that singers use don't notate the sounds for different vocal registers, but that is because it is redundant: the composer (if he is good) has done that for you.

    Furthermore, there are in fact good IPA guides for singers that make minute distinctions in vowel and consonant sounds. Thomas Grubb, author of A Manual of French Diction (also a pianist, vocal coach, and former student of Poulenc's baritone partner Pierre Bernac) painstakingly describes and notates 7 different 'e' sounds in French – and in fact has to add to existing IPA in order to notate them.

  5. john riemann soong said,

    May 15, 2008 @ 5:00 pm

    [quote]painstakingly describes and notates 7 different 'e' sounds in French – and in fact has to add to existing IPA in order to notate them.[/quote]

    Haha, my French phonetics textbooks used some of those extra symbols! But IIRC many are allophones and aren't minimal pairs, (besides the basic schwa, /E/ and /e/)? I assume that not marking a vowel distinction when it *is* a minimal pair is a bigger crime. ;-)

  6. Randy Alexander said,

    May 15, 2008 @ 9:33 pm

    Ben, thanks for this. Learning IPA for singing was part of what sparked my later interest in linguistics. IPA for singers is a little different in its representation. Length isn't indicated generally because that is indicated in the music. Quality is the most important aspect. Also, almost every word is sung in it's strong (citation) form; there is very little connected speech phonetic change. Diphthongs are realized with the stressed portion elongated and the unstressed portion shortened. Glides are almost eliminated: quickly tacked onto the end of a vowel before the next syllable. Stops are almost always realized as plosives. Stress is indicated, but sometimes overridden by the music.

    This use of IPA is quite widespread.

    As for the idea that the VT recording project is the first of its kind, Webster's Collegiate has human recordings for most of their words. I'm not sure how many words are recorded there, but I would guess it would be similar in scope.

  7. Sara Antunovich said,

    May 21, 2008 @ 12:26 pm

    "Length isn't indicated generally because that is indicated in the music."
    "Perhaps IPA that singers use don't notate the sounds for different vocal registers, but that is because it is redundant: the composer (if he is good) has done that for you."

    I'm going to have to disagree with both of you here. What Ben is referring to in terms of vowel length is, for example, in German open vowels are short and closed vowels are long, which is something that very specifically determines the word. This is something the singer does on her own. One must know how quickly the consonant comes/how long or short the vowel is; whether it is in Puccini shortening the ia in "piaccion" for the stop of the cc and staying in time, or for someone like Strauss who in his songs did not equate vowel length and note length, his "shlug ihm" here we have short open vowel and followed by a closed long vowel no matter what the note. Unless the composer is of the likes of Puccini, Verdi or Janacek, rendering the language accurately is not the priority (Janacek is a great example, he would eavesdrop on his countrymen in cafes and study the language meticulously; his operas are masterpieces of the czech language). But even with their help it is the singer's responsibility to be accurate. It is even *more* important to sing natively with composers who composed *for* their languages.

    "there is very little connected speech phonetic change….Glides are almost eliminated: quickly tacked onto the end of a vowel before the next syllable. Stops are almost always realized as plosives. …Stress is indicated, but sometimes overridden by the music."

    Here it really depends on your training, and on the rep how you make these kind of choices, everyone can make them differently. Connected speech is absolutely important for the meaning of the sentence; the complete elimination of glides and stops I don't think anyone would sanction since the rendering of every word clearly is the goal; and stress is one of the most important parts of language work. In my training I sing it as it is spoken, not for syllable to syllable clarity, but with the meaning of the entire sentence. Therefore in Italian, how closed or open a vowel is is determined by the vowels around it; that in order for a single word to sound like a single word, the vowels have to sound even, if one is very open and one very closed it sounds like 2 words or some kind of mishmash. There has to be clarity for that over the entire phrase, with stresses on where I am going in the sentence. For example, Mozart in my experience often did not write with the direction of the sentence in mind (Italian was not his native language, of course). In both arias Come Scoglio and Dove Sono, the syllables that need emphasis are sometimes on short notes. The singer/I must be able to lengthen and emphasize the important syllables without going outside the music. And let me tell you that this is hard!

    This is all to say that singers must not leave anything out. Native speaking is the goal absolutely, which is the only way, the most effective way to communicate meaning.

RSS feed for comments on this post