Timur Baytukalov has created what looks like a useful site for language learners, EasyPronunciation.com; he says:
I created this website with phonetic transcription converters – https://easypronunciation.com/en/. They can convert text into IPA phonetic transcription. I already support seven languages (English, Russian, French, Spanish, Chinese, Japanese, Italian). Russian and French converters have embedded audio recordings.
Some levels are for paid subscribers, but basic levels are free; it looks worth checking out.
I’m not sure this is especially useful for some of those languages. Russian and Spanish spelling, for example, are “phonetic” enough that anyone who has really decided to learn those languages will be able to determine the general pronunciation of a word almost immediately without the aid of a tool like this. Still, the detail that word-final -ся in Russian verbs is [sa] and not [sʲa] is a detail I would have appreciated knowing very early on in learning the language.
But developing a good accent in a foreign language is all about distinctions finer than what IPA would typically represent, either subtle gradations between what foreigners would think are the “cardinal values” of IPA vowels, or allophonic variations. For example, in Transylvanian Romanian stressed /e/ is typically realized as [ɛ], while in Moldavian Romanian several vowels are higher than the standard orthography indicates. So, if you want to speak the language like actual speakers of it instead of sounding like a foreigner who learned naively out of a book, these are details you need, and listening to audio material is much more vital than reading a very general IPA transcription.
But developing a good accent in a foreign language is all about distinctions finer than what IPA would typically represent
Oh, sure. But some of us are content with a halfway decent accent as long as we know we’ve got the basics right, at least in languages we’re unlikely to ever use in the wild.
What’s the consensus on something like forvo.com, where native speakers pronounce the words? It relies on users to pronounce the words, so languages with more speakers naturally have more words available than languages with fewer speakers.
Don’t know if there’s a consensus, but I’m kind of leery of it. You don’t know what you’re getting.
Which is true. They do have users mark their place of origin, so I can tell which French speakers are (allegedly) from France and which are from Quebec.
Poking around the English words I didn’t notice anything off, but small sample size and all.
ME: I can tell which French speakers are (allegedly) from France and which are from Quebec.
In that case, age would also be a source of differences, as well as more finely tuned regional origin.
Christopher Culver,
I am the creator of the phonetic converters.
I completely agree with you about Spanish, but not about Russian. The algorithm of the Russian converter is far from being simple. In our language, we have a huge amount of exceptions to general pronunciation rules. And there’s another thing that makes Russian pronunciation complicated – the stress position in words. It’s almost impossible to predict, especially for a beginner.
Yes, that’s why texts for students with the stress marked are so useful.
But developing a good accent in a foreign language is all about distinctions finer than what IPA would typically represent…
You could probably represent most of those distinctions using diacritics and/or you could use written descriptions, e.g., “the Romanian /i/ isn’t as close to cardinal vowel 1 as the French /i/.” I just pulled that out of my rear end, but it’s probably true, as the French /i/ is very close to cardinal vowel 1. But, yeah, I agree that you need to listen to native speakers of the language at some point if you want to have a good accent. I wouldn’t recommend trying to learn how to pronounce a language entirely from books.
You could probably represent most of those distinctions using diacritics
Or CanIPA.
I wouldn’t recommend trying to learn how to pronounce a language entirely from books.
Indeed, but sometimes there is no other way.
When it looked like I was going to be working on a language of Northern BC, I found that Franz Boas had studied it for a while in 1894 and published a volume of texts. I was able to borrow the book from a library and set out trying to sound out the written words. In spite of Boas’ key to his phonetic symbols, I found it very hard to pronounce the words. Once I was in touch with native speakers though, almost all the difficulties disappeared! Only a few unusual consonant sequences took more than a few weeks to master them.
On the other hand, when starting to study the Alsea language, extinct in Oregon, from texts published by Leo Frachtemberg in 1920, there were no speakers to listen to and thus ease my attempts at pronunciation.
And there’s another thing that makes Russian pronunciation complicated – the stress position in words. It’s almost impossible to predict, especially for a beginner.
Yes, indeed. Down with Пётр Вели́кий, destroyer of accents!
Indeed, but sometimes there is no other way.
Of course, yeah
But think about how far our accents are away from those of native speakers of modern languages when we learn how to pronounce those languages from books. How much further away must our accents be from those of native speakers of Old English, classical Greek and Latin, Alsea, etc.? There are also some things you can’t possibly know from writing alone, like intonation and voice quality. And those are a big part of what give languages and accents their distinctive sound, as we know from listening to modern languages and accents. Even if you find written descriptions of those things from native speakers of the language in question, we know from our own experience that those are often unreliable. I’ve heard native speakers of English describe their falling tones as rising tones, for example.
I’m not saying this is a big deal; it’s not to me or most people. I just get a little irritated sometimes when I hear Old English scholars, for example, claim that they know _exactly_ what Old English sounded like. They don’t. That’s like claiming you know exactly what a Trinidadian accent sounds like even though you’ve never heard one before and have only read papers about it. But anyway…I just wanted to get that last part about certain Old English scholars off my chest. The rest of this comment was really just leading up to that
Yes of course, Larry!
A few years ago while in France I saw a TV program in which actors had been trained to recite some French literature pieces (perhaps fables) from the 17C, using what was supposed to be the pronunciation of the time. Many of the words were almost unrecognizable, and from the little I understood it seemed that the pronunciations were typical of different periods, not all from what is known about the 17C. They had been somehow reconstructed by a person who was said to have done a lot of research, but whose qualifications were perhaps not quite what was needed for the task.
Yes, indeed. Down with Пётр Вели́кий, destroyer of accents!
Why do you say that?
Edit: The Cyrillic writing changed when I copied and pasted it into the comment box. Why is that?
Larry, I think it’s just the change to italic, which looks more like handwriting.
To my opinion, the best way to learn pronunciation is to learn songs with a native speaker. In this way a leaner gets not only pronunciation, but also intonation. The last one is known to be different in other languages. Just compare the music of sentences in English and Russian.
Hah, I’ve been living in blissful innocence, protected by Helvetica Neue on the Mac — even Arial has cursive-style lower-case Cyrillic letters in font-style italic which is just an oblique for Roman letters. Shame!
On the other hand Helvetica Neue has the same glyph for (lowercase) œ and (uppercase) ɶ, so shame on them too!
I just get a little irritated sometimes when I hear Old English scholars, for example, claim that they know _exactly_ what Old English sounded like.
Yes, me too! And those awful video clips in which scholars solemnly intone their ideas of Homer or Beowulf, carefully inserting their chosen values for vowels and tones and whatever and sounding utterly unlike any form of human speech. But then I have a friend who claimed that he had mastered Thai phonology after spending an hour or two with a native Thai speaker. People’s capacity for self-delusion is virtually infinite.
This is where I insert professor Arne Torp’s video skits with dramatic readings from different stages in the development of Norwegian. I find them both amusing and helpful. I think you’ve featured some of them before.
The Urnordisk clip is a couple of stanzas from the “Song of Attila” in reconstructed Proto-Norse. The Norrønt clip has the same stanzas in Old Norse.
Konungs Skuggsjá is a High Medieval (13th century) educational text, formed as a dialogue between a king and his son.
The 1400-tall (15th century) skit is about the sound changes of the time. It’s borderline understandable to a modern Norwegian ear — and more so with some exposure to conservative dialects. Roughly on the level of Övdalska or Faroese. Note that also the peeving father has the Central Scandinavian “thick l”.
He has a great villainous laugh.
What is that – [r]?
Retroflex flap [ɽ].
Thanks.
I see that one phonetician describes the Trøndersk realization of the phoneme as a post-alveolar lateral flap [ɺ̠]. I won’t object to that. But it’s getting awfully precise in place of articulation.
Larry: When Peter the Great issued his reforms to the shape of Cyrillic letters and decided which ones were and were not to be used in writing Russian, he also abolished the three accent marks (acute, grave, circumflex) in use at the time. Unfortunately, instead of changing to just one accent mark, as the Greek reformers did two and a half centuries later, he abolished accent marks tout court, leaving foreign learners in the lurch.
leaving foreign learners in the lurch.
And many Russian speakers as well. Nowhere is it as obvous as with TV newsreaders, Sámara, Costa Rica, and Samára, Russia, invariably come out as Сама́ра, (Athenian) Σύνταγμα as Синта́гма, Sanaá as Са́на, and so on. Whatever the stress in an original word, the Russian version gets stressed on the penult, as in Polish.
@ John Cowan: Ah, OK. That’s what I figured, but I just wanted to make sure. Thank you.
Whatever the stress in an original word, the Russian version gets stressed on the penult, as in Polish.
Except that Polish newsreaders, when in doubt, often stress foreign surnames and placenames on the antepenult or the final syllable to mark them out as foreign. Thus Almodóvar and Cortázar notoriously get final-stressed (despite the Spanish accent marks!), while Fitzgerald and Wyoming get initial-stressed. Only French proper names are always stressed correctly (with time-honoured exceptions, Eiffel being /ˈajfɛl/, etc.). English-speakers often do something similar to Polish placenames, e.g. stressing Poznań on the second syllable because, after all, it’s a foreign city so it should show an untypical stress pattern.
A notoriously tricky one for different reasons – it’s [ˈsindɛɣma]…
Why the [ɛ] for α?
I went to the site to see how it accents леса. It’s ле́са́. I think in our age of constraint grammar and RNN’s, it is possible to disambiguate at least a fair amount of the words involved.
P.S. But when the stress should be final, we stress the penult. Thus, we agree with the Anglosphere in misstressing Gibraltar. Such is Human Perversity.
Google translate gives Мне хочется выходить из ле́са. and Ле́са разрушаются.
Say, hardly the paragon of NLP excellence that I had imagined.
I have no idea, but apparently it’s regular somehow.
I’d want to see some backup on that (“citation needed,” as they say). I can’t think of any reason why Σύνταγμα would have [ɛ].
Piotr: Only French proper names are always stressed correctly (with time-honoured exceptions, Eiffel being /ˈajfɛl/, etc.)
The exception being justified by the fact that Eiffel is known to be a German name. English too says the /ˈajfl/ Tower. /ɛfɛl/ is the French pronunciation, conditioned by the spelling.
To be sure, the Greeks weren’t entirely faultless either. Dropping the accent mark altogether on barytones (words stressed on the antepenult, or initially if they are too short) would have saved a lot of ink and (psychological) stress. Spanish has it right: invent sensible defaults and mark the exceptions.
Why the [ɛ] for α?
This online dictionary has [ˈsindaɣma], with [a]; not [ɛ].
Just as I thought.
I’ve been there (at least to the metro station) and heard it with [ɛ]; I’ve also read something about that fact, so I’m not the only one, though it’s been a few years and I have next to no hope of finding where that was.
I’m sorry Jimmy Ho is no longer around (come back, JH!); he would have explained the whole thing. Or for that matter Nick Nicholas (come back, Nick!).
I’ve been there (at least to the metro station) and heard it with [ɛ]; I’ve also read something about that fact, so I’m not the only one, though it’s been a few years and I have next to no hope of finding where that was.
Sorry if it looks like I’m picking on you, but it’s exactly from riding the Athens subway that I have the most vivid memory of how Syntagma is pronounced. What I found most surprising back then (about 15 years ago) was that there was no audible [n] and that it sounded to me as if the word had final stress. I didn’t hear any [ɛ]. Fortunately, we live in times where people put all kinds of stuff online, so I don’t have to rely on my memory. You can hear the name of the station here twice, around the 1:29 mark and around 1:33. It’s exactly like I remembered, especially the second time with what sounds like final stress, but I’d be interested in knowing what you hear. I’m quite bad at hearing things, which makes the phonetics part of linguistics quite frustrating for me.
I’m not sure if the [n] is there (the quality of the recording isn’t great), but the stress is initial, and the middle vowel is an unambiguous [ɛ], different from the final [a].
Initial stress for sure, but I don’t hear the middle vowel as an unambiguous [ɛ]. I would have said it was [a] if I wasn’t primed to notice its quality; as it is, I can’t be sure what it is.
The stress is clearly initial on the first instance of the word, but are you sure on the second instance as well?
To me it sounds different. And while I agree that the vowel in the second syllable is different from the one in the last, to me it’s still nearer to [a] than to [ɛ].
That’s exactly what I hate about phonetics.
Edit: While I responded to David’s comment, Hat commented as well. Hat, so you hear initial stress in both instances of the word, like David?
The word definitely has initial stress, but in the second occurrence it’s said with a higher pitch on the final syllable.
Here’s the first one isolated.
http://www.hau.gr/?i=learning.en.podcasts-in-greek
In Lesson 4 Σύνταγμα occurs a few times.
Since we have tangible phonetic data, why not analyse them instrumentally, as behooves a scientist? I have used Praat to compare these vowels. Since this is an informal exercise and the quality of the recording is poor, a rather wide margin of error must be assumed. Here are the results for the medial and the final vowels in both occurrences of Syntagma, for what they are worth:
The 1st occurrence: F1=672 Hz, F2=1879 Hz; F1=641 Hz, F2=1858 Hz.
The 2nd occurrence: F1=624 Hz, F2=1841 Hz; F1=633 Hz, F2=1688 Hz.
However, the second realisation has a diphtongised medial vowel, with F2 dropping from ca. 2134 Hz to 1562 Hz. No such effect is visible in the first realisation.
I have also measured the formants of the second vowel of Panepistimio (another Metro station on the same line, mentioned at 70-71.5 s):
F1=473 Hz, F2=2240 Hz.
Discussion: the first a of Syntagma is realised sligthly differently from the second, but the difference is probably due to its weak rhythmic position. All four measurements are consistent with a slightly raised and slightly fronted [ɐ] vowel (comparable to Australian English STRUT), still more fronted towards [ɛ̞] when found in a rhythmic dip. In the second realisation of Syntagma the first a moves from a front to a backish position (while remaining near-open). The back offglide may result from assimilation to the following /ɣ/. The second vowel of Panepistimio is definitely higher and fully front (more or less halfway between cardinal [ɛ] and [e]).
I missed the second occurrence completely.
Same middle vowel, same impossibility to tell if the n is there or nasalizes the preceding vowel or is completely gone. The pitch on the last syllable is so high (I’m sure there’s a comma after it) that it may not make sense to ask whether the first or the last syllable is more stressed – which isn’t what I expected from a language with phonemic stress! This kind of thing happens in French all the time, but not in German.
Piotr, thanks for the idea about comparing to e elsewhere. That’s indeed a bit different, namely higher; and the one in both occurrences of Syntagma isn’t quite what I think cardinal [ɛ] is, but it’s much closer to that than to my somewhat vague idea of what cardinal [æ] is. I have next to no experience in hearing centralization, so thanks for the measurements! I really should learn how to use Praat at some point.
but it’s much closer to that than to my somewhat vague idea of what cardinal [æ] is.
I wrote [ɛ̞], with the “lowered” diacritic, which is roughly the same species as [æ] (not regarded as a cardinal vowel). The first /a/ in the second realisation does start close to [æ], but quickly glides towards its back counterpart (the decrease of F2 is practically linear in time).
Dziękuję bardzo!
I will check the nasal when I’m back home from work.
I’ve had a closer look at the first realisation of Syntagma (the second has a lot of background noise superimposed over the first syllable). The total duration of the word is ca. 830 ms. The first 100 ms is indeterminate fricative noise (the quality of the spectrogram is so poor that the expected energy peak above 5000 Hz for /s/ is not visible). Then we have a relatively high front vowel (ca. 70 ms long), then there’s a nasal visible in the spectrogram (the F2 of the vowel is attenuated and a clear nasal resonance appears at ca. 2500 Hz), lasting ca. 60 ms, and then a stop (ca. 45 ms). There is no sharp boundary between the /i/ and the /n/, but they are quite clearly distinct as segments, so yes, the nasal is definitely there. Fine details like the degree of vowel nasalisation are drowned by interference.
David M: it may not make sense to ask whether the first or the last syllable is more stressed – which isn’t what I expected from a language with phonemic stress! This kind of thing happens in French all the time, but not in German.
Having spent a much longer time than usual in France this year, I listened to a fair amount of radio programs there. One thing that struck me was the number of professional speakers – announcers, reporters, and some interviewees – who stressed the first syllable of words, as for instance: le NOUveau PRÉsident a DÉclaré … I pointed this out to my sister (with whom I was staying in the Paris area) but she did not consider it particularly noticeable – I guess she was quite used to it. This new heavy stress caused initial consonants to be uttered quite forcefully but otherwise unchanged from the normal (no aspiration, for instance, as with an English or German accent), and the vowels were not affected either.
Weird!
Yes, and I found it tiring to listen!
David: the difference between Ultimate [ɛ] and Ultimate [æ] is that in the former, you drop your jaw to your chest, whereas in the latter, you both drop your jaw to your chest and stick your tongue out and lick your chin. In other words, [ɛ] is -ATR whereas [æ] is +ATR.
Marie-Lucie: I hear an analogous thing in American English announcements, whereby comparatively unimportant words get sentence stress: “Ladies and gentlemen, we are going to BEGIN boarding AT this TIME.”
JC, in the French case, les MOTS IMportants get initial stress.
Pre-boarding words: often the acoustics in airports are so bad that it is very hard to understand the announcements. That may be why the personnel needs to emphasize even normally unimportant words.
Perhaps French is re-evolving lexical stress domains after centuries of treating stress as a purely phrasal phenomenon. See here for a phonological analysis of tonal phenomena in French accentual phrases (including “over-accented” ones). An auxiliary tone occurring naturally at or close to the initial phrase boundary can easily be co-opted for demarcative functions (instead of marking the beginning of an accentual phrase it marks the beginning of important words). Hasn’t l’accent affectif been the norm for a long time in emphatic expressions like c’est DÉgoûtant or quel homme aBOminable?
Something similar is happening in Polish too. Polish stress is normally penultimate, but in longer words there is also a secondary stress on the first syllable. If a word is emphasised in public-speaking styles, the relative strength of the two stresses may be reversed. A comedian parodying a ponderous political speech would almost certainly make extensive use of this initial-stress thing. If this trend continues, Polish will go the way of Czech.
Oh yeah, I forgot about the special status of [æ]…
Ha, that was exactly my impression!
Yes, that’s been normal for professional speakers for quite a while; I think I mentioned it here a few (or not so few) years ago. The phrase-final stress is still there as far as I’ve noticed, but the word-initial high pitch is at least as prominent.
Now that I think of it, this is also normal in dégeulasse, but not in dégeu.
…which would be Total Polonic Reversal.
We’ll have to do a Polonoscopy to be sure.
Merci Piotr et David!
Yes, perhaps this initial stress is a generalization of l’accent affectif.
David: this is also normal in dégeulasse, but not in dégeu.
You mean déGUeulasse and its short form déGUeu (‘disgusting, puke-making, etc’), both ‘affective’, slangy words. (GU here is [g|).
The root of dégueulasse is la gueule, literally ‘carnivore’s mouth’, hence (slangy) ‘mouth’, and the immediate stem is the verb dégueuler (slangy) ‘to puke, to upchuck’. The suffix -asse on an adjective is usually derogatory.
I’m pretty sure David knows all that and just typed it wrong. French is tricky with all the -eu- and -ue- and silent vowels.
Whatever the stress in an original word, the Russian version gets stressed on the penult
The only exception that comes to mind is Malibu, which is always Малибу́.
LH: I’m pretty sure David knows all that and just typed it wrong. French is tricky with all the -eu- and -ue- and silent vowels.
Yes, but David is not the only one who will read my comment.
@JC, +ATR /æ/ is the American version, innit? The near-RP I lernt in school has no +ATR I think.
+ATR in words like gade is one of the shibboleths for (central and northern) Swedes trying to speak Danish — they don’t have it and don’t hear it, while in Danish it’s the main distinction between the phonemes /æ/ and /ɛ/.
Oh! Yes, thanks, marie-lucie.
Yes, it’s North American /æ/ that has +ATR, but [æ] in square brackets was defined when both North American and English accents used it as their TRAP vowel. The English TRAP vowel has moved since, and some varieties of American have diphthongized TRAP in some or all cases, but that of course doesn’t affect the meaning of [æ].
From what I understand, Standard Finnish continues to have [æ] for written ä, short or long.
[æ] was not originally defined as [+ATR], only as near-open (between cardinal [ɛ] and [a]). It is true, however, that the typical RP realisation at the time was [+ATR], or as Daniel Jones put it in An outline of English phonetics,
Still, the symbol is often used for near-open front vowels in various languages and is not reserved for [+ATR] vowels only (see the IPA chart).
Except that “near-open” is an articulatory lie; I can pronounce bet as [bɛʔ] with my mouth as open as my jaw articulation will allow. It ends up sounding pharyngealized, but is clearly distinct from bat [bæʔ] pronunced in the same way. (I have to use [ʔ] because with my mouth fully open my tongue cannot reach my alveolar ridge or upper teeth; the frenum is too short.) In essence: [ɑ], [ɛ] (or [ɛˤ] if you insist) and [æ] are all fully open vowels with +RTR, neutral TR (that is, -RTR -ATR), and +ATR respectively.
OK, I’m talking of traditional definitions (as perpetuated in IPA’s own charts), not about what IPA transcription should be like if it were truly based on scientific principles.
I can pronounce bet as [bɛʔ] with my mouth as open as my jaw articulation will allow. It ends up sounding pharyngealized, but is clearly distinct from bat [bæʔ] pronunced in the same way.
Hmm, since when is vowel height defined with respect to the jaw and not the tongue body? [ɛ] with the jaw fully open is certainly doable, but seems to require a bit of added tongue-bunching to maintain the constriction height. (I find even an [e]-ish vowel is vaguely reachable in this way.)
Bringing [ɑ] to the mix seems unlikely to clarify things much either, since this now adds also backness as a confounder. And the same goes even if you meant [a]; if one distinguishes cardinal [æ] — [a] — [ɑ] by frontness or by tongue root retraction feels to me like more of an issue of analysis than any real articulatory difference. The tongue has to go somewhere, and in a maximally lowered position, any tongue root advancement will most naturally translate into tongue blade fronting (and vice versa, and the inverse). True, one can do some additional stretching or bunching bathybatics to fill up the entire 3×3 [±front ±back ±ATR ±RTR] grid if desired, but that surely doesn’t imply that we would have to take the tongue root neutral low vowels as “the” cardinal ones.
Hmm, since when is vowel height defined with respect to the jaw and not the tongue body?
When I do the Extreme Low Vowels, my tongue body is directly on the floor of my mouth, so they come to the same thing.
bathybatics
???
Neither Google nor the OED knows any such word.
“the” cardinal ones
Being self-taught, I don’t actually know how to execute “the” cardinal vowels.
bathybatics
acro- ‘high”
bathy- “deep”
acrobatics – bathybatics