Vakgroep Linguïstiek, Trans 10, 3512 JK Utrecht, telefoon 030-2536265, fax 030-2536000

Maya van Rossum

eindwerkstuk cursus TEKST-NAAR-SPRAAK-SYSTEMEN 1996-97


"De positie van het Afrikaans staat in het nieuwe Zuid-Afrika ter discussie. Het is weliswaar een van de elf officiele talen, maar heeft zijn dominante positie verloren." - (Elsevier)
The first Dutch colonists arrived at the Cape of Good Hope during the seventeenth century, to establish a "halfway station" responsible for fresh supplies to NOIC vessels en route from the Netherlands to the East.This Dutch speaking minority not only brought their Malay slaves with them, but also developed "trade" relations with the neighbouring African nations.Very soon traces of these other languages could be found in the spoken Dutch of that time, words such as bredie and bobotie (Malay food dishes), piering (saucer), donga (hole in the ground), shongololo (centipede), etc. became part of the "accepted" vocabulary. Over the next two centuries, there was an influx of French, German and British settlers and each of these languages left their mark on the colonial Dutch, to such an extent that a new dialect, Afrikaans, emerged, which co-existed uncomfortably with the official High-Dutch. This new "kombuistaal" (kitchen language) was not acknowledged and certainly not spoken or heard in sophisticated society. With time however, knowledge of spoken "High-Dutch" faded and was replaced by Afrikaans: a Dutch relative, but molded by European and African languages.

This process was and is obviously not unique; in a similar way, Yiddish, the vernacular of the old Eastern European Jewish community is a fusion of Middle High German with Old French, Old Italian, Hebrew/Aramaic and the Slavic languages (Alpert, 1995).

From personal observation it can be said that there still exists a very strong likeness between some Yiddish words and German: with a reasonable knowledge of German, one can easily pick out and understand at least a small amount of Yiddish, if spoken or sung fairly slowly and clearly. When presented with the printed text, the spelling is very different, though.Of course Yiddish was forged over the course of a thousand years, whereas Afrikaans is a much younger language; not yet three hundred and fifty years old. It would therefore seem reasonable to expect a much stronger link between Afrikaans and its ancestor Dutch, both in the spoken and written form.
Recently, a preliminary Dutch diphone-set, which consists of 2400 diphones, spoken by a native Dutch speaker, was developed for MBrola, a diphone-synthesiser, (developed by T. Dutoit and V. Pagel at Faculte Polytechnique de Mons).Considering the strong connection between Afrikaans and Dutch that has been proposed earlier, it should be possible to synthesise Afrikaans utterances, using the Dutch Diphone set and MBrola-synthesiser. (This particular set is not of studio quality, but a new, studio quality set should soon be commercially available from Fluency Speech Technology.)


Choosing the Afrikaans speech material to be synthesised, two considerations were taken into account: the utterance to be synthesised should be a fair representation of the different sounds that exist in the Afrikaans language but it should also be contextually relevant and interesting. For instance, synthesising "the old man is walking his dog " would be rather boring and meaningless. In the end the following utterance was assembled: (A=Afrikaans; Di=diphone symbols used; D=equivalent Dutch utterance; E= English translation. The pitch contours are also given)

A: Reen dit nog steeds so baie in die Kaap, of skroei die son buite, op Blouberg se strand,

Di: ri@n@tnOxsti@tsubAj@nikAp_OfskruidisOnb2tObl2YbEGxs@strAnt_

D:Regent het nog steeds zo veel in de Kaap, of brandt de zon buiten, op Bloubergstrand

Di: reG@nt@tnOxstetzovelInd@kap_OfbrAnt@zOnb9yt@nOblAubERxstrant

E:Rains it still so much in the Cape, or scorches the sun outside on Bloubergbeach

A: Ons mis die wyn, braais en die heuwels vol proteas.

Di: O~sm@siwein_braisInijY@w@LsfOLprutiAs

N: We missen de wijn, barbecues en de heuvels vol proteas

Di: w@mIs@d@wEinbArb@kjuzEnd@h 2v@LsfOLproteas

E:We miss the wine, barbecues and the hills covered in proteas

The general guidelines on segment durations mentioned by Klatt (1976) were followed, except when, while listening to the resulting utterance, some adaptations seemed necessary. The pitch contour was added solely "by ear": no intonation grammar exists for Afrikaans as it does for Dutch (´t Hart et al. 1990).The biggest problem lay in finding the correct diphones for the Afrikaans vowels and diphthongs, as these are produced very differently to their Dutch counterparts. The procedure used to find the diphthongs, was as follows: first the phonological descriptions of the Dutch (Trommelen & Zonneveld, 1989) and Afrikaans (personal communication, M.A. Kemp, Dept. of Linguistics, University of Stellenbosch) were compared and subsequently, those Dutch vowels which were the closest alternatives to the Afrikaans diphthongs, were synthesised. One could therefore speak of three columns or streams: 1. the Dutch diphthong; 2. the Afrikaans diphthong; 3. the Dutch alternative used during synthesis.

Thus for the Dutch "au" (as in blauw) the movement is from /A/ = low, back towards /o/ = middle, back, rounded.The Afrikaans "ou" is formed by a movement from a mid, back, rounded position to high, back and rounded. The closest approximation would therefore be the Dutch /o/ and /u/. The Dutch "ei" (as in wijn) can be seen to move from the /E/= mid, front, towards /I/= high, front. The Afrikaans equivalent is produced by moving from a mid, front position to a high front position; similar to the Dutch /e/ to /i/. Lastly, the Dutch /ui/ consists of a movement from /^/= a mid back position towards the /y/= a high rounded front vowel. In Afrikaans this would be produced by starting at a mid rounded front position and moving to a high rounded front position.This could therefore be represented by the Dutch /o/ and /y/. To establish how accurate these alternatives were, the three versions of each diphthong (the Dutch, Afrikaans and Dutch alternative) were recorded by me as I am fluent in both Dutch and Afrikaans, with a DAT-recorder in a soundproof room, taking care to maintain the same distance to the microphone.These were then saved as sound files and via a script written especially for this type of analysis, the formant tracks were obtained in the programme "praat". To see how accurate this analysis was, the formant motions of one of the Dutch diphthongs were used to specify formant frequency control parameters and this information was used to synthesise the /au/ with the Klatt-synthesiser (Klatt, 1979). Synthesising a female voice on the Klatt- synthesiser is not ideal and the decision was made to multiply the formant values by 0,9 to approximate a male voice.


Looking at, and listening to the Afrikaans utterance [with its command file], many observations can be made, some of them more interesting than others. Comparing the Dutch and Afrikaans scripts, there seem to be remarkable similarities: the symbols and placement of the consonants are virtually identical, except that the /z/ does not exist in Afrikaans, /x/ is sometimes replaced with /k/ as is the /tj/. Most words are the same, except, for instance, "veel" for "baie" and "barbecue" for "braai". This is quite different to the Yiddish/German link mentioned in the introduction, as the Yiddish spelling seems to deviate more but the pronunciation less (eg. "ikh" for "ich" where the only difference is that the Yiddish fricative is a bit more strident; "arayn" for "herein" the only difference being the omission of the /h/). The most interesting and main difference between the Dutch and Afrikaans seems to be in how the vowels and diphthongs are formed and it might have been useful to compare these differences in terms of an articulatory model (movement of articulatory parameters in the time domain) as this would automatically have given the acoustic consequences, and choosing the Dutch alternatives would have been less guesswork and more theoretically founded. Unfortunately, this type of data is as yet not available.

In the introduction it was claimed that there should be a connection between the Afrikaans and Dutch. What better way to test this than to present the Afrikaans utterance to a native Dutch speaker with no knowledge of Afrikaans; the reaction of the Dutch subject was illuminating: "Ik kan hier ECHT NIETS van bakken." which roughly translates into the fact that it might as well have been Russian or Marsian, for all he could tell.One can interpret this outcome in different ways:1- modern spoken Afrikaans has diverged to such an extent that it has very little in common with its ancestor; 2- as its creator, I might find this perfectly acceptable Afrikaans, whereas, in reality, this utterance might not have any resemblance to the real thing, or to any other known language, for that matter; 3- this utterance was synthesised at a normal speaking rate, including the processes of assimilation, deletion and reduction which one would expect and can therefore not be compared to the example of Yiddish mentioned in the introduction (slow rate, clear speech). The second deduction was easily, though expensively, eliminated. The Afrikaans utterance was presented over a speaker, through the telephone, to a native Afrikaans speaking South African. The subject was able to repeat the utterance verbatim. An English speaking South African was asked to write down what he heard, this also proved to be no problem. The utterance was indeed Afrikaans, as it was meant to be. The first assumption also does not hold. I have witnessed Dutch immigrants and holidaymakers newly-arrived in South Africa, without any previous knowledge of Afrikaans, coping remarkably well when forced to follow Afrikaans instructions or conversation, provided that the Afrikaans was spoken very slowly and clearly. This immediately confirms the third possibility and it might be that if a Dutch subject was provided with the written script, understanding improves. To test this, a second native Dutch subject was asked to listen to the Afrikaans utterance while following the written text and she found it quite easy to do so. The Afrikaans synthesised with the Dutch diphone system seemed to be acceptable Afrikaans. The Dutch alternative diphthongs seemed to have been perceptually irrelevant. It must be remembered however, that the rate of speech, the context provided by the rest of the utterance and maybe even the neighbouring consonants might have masked the slight differences that could exist.

This brings us to the three versions of the diphthongs that were recorded for comparitive analysis. It proved to be rather difficult to produce these naturally in isolation as it involved switching from one language's mindset and "mouthset" to another and producing the Dutch alternatives without lapsing into an Afrikaans mouthset for which it would be natural.The Dutch diphthongs were again presented to a Dutch subject who only commented that they were rather long but otherwise quite acceptable. This is probably a result of the fact that they were spoken carefully and in isolation.

Dutch vowels: au, oo, ei, ee, ui, eu.
Afrikaans vowels: au, ei, ui.

There seemed to be nothing very much wrong with the formant tracks, however (see appendix). If one theoretically plotted the movements of the diphthongs in terms of F1 and F2, they could be described as follows (see ´t Hart, 1994):
For the Dutch "au"- highish F1/lowish F2 to midlow F1/low F2. For the Dutch alternative/ approximation to the Afrikaans-midlow F1/low F2 to very low F1/very low F2. It is clear from the formant tracks that the F2 of the Afrikaans deviates; it would seem that the Afrikaans /o/ lies further to the front, more like the English blow.
The Dutch "ei"- highish F1/mid high F2 to lowish F1/high F2. The Dutch alternative-low F1/high F2 to very low F1/very high F2. Comparing this to the Afrikaans "ei"-formant tracks, the Afrikaans has a higher F1 and a lower F2 = lower tongue position and further to the back.
The Dutch "ui"- highishF1/mid F2 to very low F1 /high F2. For the Dutch approximate- lowish F1/ highish F2 to very low F1/high F2. This was also approximately what the Afrikaans version looked like (see appendix for drawn formant tracks).

We have seen that the Dutch approximates, when synthesised in an utterance, are perceptually acceptable, even though there exist some differences when comparing the formant tracks. These differences can be clearly heard when listening to the recorded diphthongs produced in isolation. The Dutch "au" that was synthesised with the Klatt-synthesiser, was also presented to a native Dutch subject. She found this a more acceptable Dutch diphthong than the recorded version, even though the same (0,9 x) formant values were used. This can be explained quite easily: I had shortened the duration of the diphthong and it was now perceived as more natural [listen to the soundfile, or look at its input script]. As there are more then one acceptable pitch contours possible for the Dutch utterance and since the Afrikaans utterance is only a very small sample, a comparison between the Dutch and Afrikaans intonation patterns seemed inappropriate and was not attempted.


A striking feature, when listening to, and looking at the Afrikaans utterance [with its command file], is the extensive and repititive use of assimilation (for example "in die" becomes "@ni") and reduction to schwa whenever possible. This simplification can be found on other levels as well. The spelling, for example, of a lot of words compared to the Dutch equivalent: "regen" becomes "reen" -and there are thousands of these examples: "echt" becomes "eg"; "nieuw" becomes "nuut"; "theorie" becomes "teorie". The same holds when it comes to the use of verbs, Afrikaans has only one verb for single or plural form: ek eet / ik eet; jy eet /jij eet; ons eet / wij eten; hulle eet / zij eten; almal eet.

Should Afrikaans be thought of as a simplified version of Dutch, maybe even a bit "backward", the type of language a child would produce before it is fully able to cope with the complexities of its mother tongue? This is a popular view held by many Dutch visitors to South Africa. But Afrikaans is perfectly sufficient as a language and has been for approximately three hundred years. It seems that this language has achieved a remarkable simplicity and uniformity, both in grammar and in speech, stripping it of all but the necessary for successful communication. It reminds one of a similar trend seen in the music of some American composers (Michael Nyman and Philip Glass), the so-called minimalists (Griffiths, 1993). In fact one could describe Afrikaans as a minimalist language and even maybe the most progressive, sophisticated of the germanic languages.

Whatever one's point of view may be, there still exists a strong link between Afrikaans and its ancestor, Dutch, even though there are some surprising differences between the languages.


aanvang document
Universiteit Utrecht Faculteit der Letteren Uw reactie

Laatst gewijzigd: 17 maart 1997 (MvR, HQ) / Hugo Quené