Content uploaded by Gilles-Maurice de Schryver
Author content
All content in this area was uploaded by Gilles-Maurice de Schryver on Apr 25, 2014
Content may be subject to copyright.
Towards Strategies for Translating Terminology into all South African
Languages: A Corpus-based Approach
Rachélle GAUTON°, Elsabé TALJARD‡ & Gilles-Maurice DE SCHRYVER#
Department of African Languages, University of Pretoria, SA° ‡ # &
Department of African Languages and Cultures, Ghent University, Belgium#
1. Introduction
The single biggest problem that translators who translate from a language such as
English into the African languages have to contend with is the lack of terminology in
the African languages in the majority of specialist subject fields. The relevance of
terminology theory and practice for translators therefore becomes clear when the
translator is faced with a situation where he/she can no longer rely on existing
knowledge and/or dictionaries, and has to conduct research beyond the dictionary.
There is a clear difference between translating into an international language
such as English and translating into so-called ‘minor languages’ or ‘languages of
limited diffusion’ (LLDs) such as the African languages. This difference also holds
regarding the translation of terminology. Cluver (1989: 254) points out that since the
terminographer working on a developing language actually participates in the
elaboration / development of the terminology, he/she needs a deeper understanding of
the word-formation processes than his/her counterpart who works on a so-called
‘developed language’.
In this paper, a preliminary study is undertaken, comparing and analysing the
various translation strategies utilised by African-language translators in the finding of
suitable translation equivalents for English terms foreign to the African languages. To
this end, a multilingual corpus of ten parallel texts in all eleven of the official South
African languages has been studied. These parallel texts have been culled from the
Internet, and a full report on the building of this multilingual corpus can be found in De
Schryver (2002). The combined size for all eleven parallel corpora is 348,467 running
words, or thus nearly 32,000 words on average per language.
2. Methodology followed
The first step in this pilot study is to extract the relevant terminology and to compare
the English terms with their translation equivalents in the nine official African
languages, viz. isiNdebele, siSwati, isiXhosa, isiZulu, Xitsonga, Setswana, Tshivenda,
Sepedi and Sesotho, as well as with Afrikaans. For the purposes of this study, we
assume that the English texts are the source texts, as all of the websites from which the
parallel texts were downloaded, have been written in English, with only small selected
sections of the sites in question being provided in the other official languages.
Furthermore, it is standard practice in South Africa when undertaking a translation
R. Gauton, E. Taljard & G-M de Schryver — Translating Terminology into all SAn Languages 81
project involving all nine of the official African languages, to provide the source text in
English, as this is in the majority of cases the only language that all of the translators
have in common. This is especially the case when the subject matter of the text in
question is of a technical nature, as the African languages do not as a rule possess the
requisite terminology. On the basis of this evidence it is therefore highly unlikely that
any of the African languages (or for that matter Afrikaans) would have served as the
source for the texts culled from the Internet on which this study is based.
In extracting the terminology from this corpus of parallel texts, the methodology
illustrated by Taljard & De Schryver (2002) is followed. These researchers have shown
how African-language terminology can successfully be extracted semi-automatically
from untagged and unmarked running text (texts culled from the Internet are, when
saved as text files, an example of this) by means of basic corpus query software like
WordSmith Tools. The key procedure for identifying terminology in each of the
parallel corpora is to compare the frequency of every distinct word-type in each
parallel corpus, with the frequency of the same word-type in respective reference
corpora – the reference corpora obviously being the bigger of the two in each case.
Items displaying a great (positive) disparity in frequency are identified as terminology,
since the disparity would imply that those specific items occur with unusual (high)
frequency in the smaller corpus. The terminology retrieved in this way across the
parallel corpora compares very well (see also Uzar & Walinski 2000). For the purposes
of this study, the eleven general corpora compiled in the Department of African
Languages at the University of Pretoria have been used as reference corpora (for more
details, cf. Prinsloo & De Schryver 2002: 256). Sizes of these corpora are typically
from a few million up to more than 10 million running words each.
The next step is then to identify the various translation strategies utilised by the
different translators in finding suitable translation equivalents for the English terms
identified.
3. Preliminary results
On studying the various outputs from the keyword searches done using the two sets of
eleven corpora, i.e. the eleven parallel corpora versus the eleven general corpora, the
following is readily apparent:
• Although the number of keywords thrown up semi-automatically differs from
language to language, there is a good correlation across the parallel corpora
between the terms obtained in this manner.
• Even at a casual glance, the following strategies utilised in the translation of source
text (ST) terminology are immediately obvious:
Translation by means of loanwords in which the English spelling has been
retained. Such words have not been transliterated, i.e. nativised in the sense
that their phonology has been adapted to reflect the phonological system of
the borrowing language.
TAMA 2003 South Africa: CONFERENCE PROCEEDINGS 82
Term formation through transliteration. New scientific and technical terms are
formed via a process of transliteration by adapting the phonological structure
of the loanword to the sound system of the borrowing language.
The occurrence of these two translation strategies in the various languages is
summarised in Table 1.
Table 1: Keywords and the translation strategies pertaining to loanwords in eleven
parallel corpora
Language Keywords Loanwords with
English spelling
Transliterations
# # % # %
isiNdebele 583 14 2 37 6
siSwati 427 18 4 16 4
isiXhosa 580 27 5 32 6
isiZulu 619 71 11 30 5
English ST 443
Afrikaans 426 10 2 17 4
Xitsonga 402 37 9 56 14
Setswana 436 26 6 32 7
Tshivenda 394 43 11 55 14
Sepedi 371 18 5 32 9
Sesotho 320 10 3 13 4
The most important findings regarding these two translation strategies are:
• Whereas isiZulu seems to make use of non-nativised loanwords to a larger extent
than transliterations, and whereas siSwati uses these two strategies in equal
measure; in all the other languages, i.e. isiNdebele, isiXhosa, the Sotho languages
(Setswana, Sepedi and Sesotho), Tshivenda, Xitsonga and Afrikaans,
transliterations seem to be used to a greater extent than non-nativised English
loanwords as preferred translation strategy for technical terms.
• Many of the non-nativised loanwords under discussion here, are in fact English
abbreviations such as SAQA (South African Qualifications Authority), NSB
(National Standards Body), RPL (Recognition of Prior Learning), etc. that have
been taken over as such into the borrowing language. In Sepedi and Sesotho for
example, 78% and 70% respectively of the non-nativised loanwords are English
abbreviations that have not been translated into the language concerned, but taken
over as is.
• In Afrikaans, translation equivalents are given for English abbreviations such as
Eng. SAQA : Afr. SAKO (Suid-Afrikaanse Kwalifikasie Owerheid), Eng. NSB :
Afr. NSL (Nasionale Standaardeliggaam), etc., with the noted exception of the
abbreviation ANC (African National Congress).
R. Gauton, E. Taljard & G-M de Schryver — Translating Terminology into all SAn Languages 83
• A similar situation is found in isiNdebele, where translation equivalents are
provided for abbreviations such as: Eng. NSB : Ndeb. iHTB (iHlangano
yesiTjhaba yamaBanga); Eng. SAQA : Ndeb. iPSAF (UbuPhathimandla
beSewula Afrika), etc.
4. An illustrative example: comparing translation strategies utilised in isiZulu
and Sepedi
As was stated at the outset, this paper is intended as a preliminary investigation into
strategies utilised in the translation of terminology into all South African languages.
As this is an ambitious and wide-ranging project, and as time is limited in a forum
such as this, two languages, viz. isiZulu and Sepedi, are used as an illustrative example
of this process. In Table 2, a representative sample of 20 SL terms are selected from
our large database currently under construction, and this is followed by a comparative
analysis of the strategies used in the translation of terminology into these languages.
Table 2: Comparative analysis of 20 SL items translated into isiZulu and Sepedi
SL term isiZulu translation equivalent Sepedi translation equivalent
accreditation PAU: ukunikezwa amandla / igunya;
ukugunyaza BT: to be given the power /
authority, security; to authorise.
MGW: netefatšo BT: verification;
MGW: tumelelo BT: permission,
approval.
agenda PAU: uhlelo / uhlu lokuzoxoxwa ngakho BT:
arrangement, list of things (issues) that will be
talked about / discussed.
MGW: lenaneo BT: list, programme.
apartheid MGW & MNW: ubandlululo (ngokwebala)
BT: discrimination (on the basis of colour),
exclusion.
LWT: aparteiti BT: apartheid;
MGW: kgethollo BT: separation,
segregation - SYN.
assessment
criteria
PAR: inqubo yokuvivinyisisa / yokuvivinya BT:
criteria (lit. procedure, process) of examining /
examining thoroughly; PAR: indlela yokuhlola
BT: manner of examining. (All of these
paraphrases are rather vague and do not succeed
in capturing the exact meaning of the SL term.)
PAR: mokgwa wa tekanyetšo BT:
way / manner of estimation; PAR:
dinyakwa tša tlhahlobo BT:
requirements of examination.
census LWE: i-census; MGW: ubalo BT: count (n) -
SYN.
MGW: palo BT: count (n.).
definitions MGW: izincazelo BT: explanations. MGW: dithlalošo BT: explanations,
meanings.
documentation MGW: izincwadi BT: letters, books; MGW:
amabhuku BT: books.
LWT: ditokumente BT: documents.
finance /
financial
MGW: izimali / wezimali BT: money / of
money.
MGW: (wa / tša) tšhelete BT: (of)
money.
gender RTE: ubulili BT: gender. RTE: bong BT: gender.
global PAU: umhlaba wonke jikelele BT: the whole
earth, world.
PAU: lefase ka bophara BT: the
world at large.
TAMA 2003 South Africa: CONFERENCE PROCEEDINGS 84
guidelines COM: imihlahlandlela BT: < -hlahla ‘guide’ +
(i)ndlela ‘way, manner’ (Note that the same
term is also used to designate ‘framework’.);
COM: imikhombandlela BT: < -khomba ‘show’
+ (i)ndlela ‘way, manner’.
COM: methalohlahli BT: < methala
‘lines’ + hlahla ‘guide’; COM:
ditšhupatsela BT: < šupa ‘show’ +
tsela ‘road, way’.
institutions SSP: izikhungo BT: (lit.) gathering places. LWT: diinstithušene BT: institutions.
Minister RTE: ungqongqoshe BT: minister. CST: tona BT: advisor to the chief /
king.
outcome(s) MGW: imiphumela BT: results; RTE:
impumelelo BT: outcome, success.
MGW: dipoelo BT: results.
redress MGW: ukulungisa BT: to correct, rectify. MGW: phetolo BT: change, reversal.
regulation(s) COM: imithethonkambiso BT: < imithetho
‘laws, rules’ + (i)nkambiso ‘custom’.
MGW: melawana BT: small laws.
research RTE: ucwaningo BT: research. LWT: resetšhe BT: research; SSP:
nyakišišo BT: investigation - SYN.
South African
Qualifications
Authority
LWE: i-South African Qualification(s)
Authority; PAR: Isigungu seziPhathimandla
sokwengamela iziqu eNingizimu Afrika BT:
authorising committee that presides over South
Africa's qualifications - SYN.
PAR: Bolaodi bja Mangwalo a Thuto
bja Afrika Borwa BT: authority of
letters of learning of South Africa.
stakeholder(s) MNW: abathintekayo BT: those affected. COM: bakgathatema BT: those who
take part.
Standards
Generating
Body
LWE: iStandards Generating Body; PAR:
uMgwamanda eKhiqiza / eYenza amaZinga BT:
assembly, congregation, community that
(abundantly) produces / makes standards; RTE
& LWE: uMgwamanda iStandards Generating
Body - SYN.
PAR: Lekgotla la Tlhamo ya Maemo
BT: council of establishment of
standards.
Note that in Table 2, the SL terms are listed as proffered by the keyword search, i.e. in
derived or inflected form. However, should a terminology list be compiled, these terms
will be lemmatised under their canonical forms. Note also that the following codes are
used to symbolise the strategies that are, according to Baker (1992: 26-42), often used
by professional translators in solving various types of problems of non-equivalence at
word-level:
• MGW: Translation by a more general word (superordinate).
• MNW: Translation by a more neutral or less expressive word.
• CST: Translation by cultural substitution.
• Translation using a loan word or loan word plus explanation (sometimes in
brackets):
LWE: Translation by means of loanwords in which the English spelling has
been retained. Such words have not been transliterated, i.e. nativised in the
sense that their phonology has been adapted to reflect the phonological
system of the borrowing language.
R. Gauton, E. Taljard & G-M de Schryver — Translating Terminology into all SAn Languages 85
LWT: Term formation through transliteration. New scientific and technical
terms are formed via a process of transliteration by adapting the phonological
structure of the loanword to the sound system of the borrowing language.
• PAR: Translation by paraphrase using a related word, i.e. paraphrasing by using a
direct / ready equivalent of the SL item in the paraphrase.
• PAU: Translation by paraphrase using unrelated words, i.e. paraphrasing by not
using a direct / ready equivalent of the SL item in the paraphrase.
In addition to the translation strategies listed above, it is well known that translators
working into the African languages are more often than not required to create new
terms, and should therefore be completely au fait with term creation strategies in their
particular language. Regarding term formation in the African languages, Mtintsilana &
Morris (1988: 110-112) distinguish between term-formation processes internal to the
language, and borrowings from other languages. They identify a number of term
formation processes in the African languages, of which the following appear in Table
2:
• Semantic transfer: This is the process of attaching new meaning to existing words
by modifying their semantic content.
SSP: In the creation of new terms, the most common form of semantic
transfer is semantic specialisation, i.e. a word from the general vocabulary
acquires an additional, more technical meaning.
• COM: Compounding. The term is coined by combining existing words.
• SYN: Synonym richness of the vocabulary. Although this is not a method of
creating new terms, Mtintsilana & Morris point out that the relative abundance of
synonyms in African-language vocabularies offers both advantages and
disadvantages from a terminological point of view. E.g., a term may be coined for a
foreign concept while a transliteration of the foreign term is also in use.
• Lastly, in some cases in Table 2 above, there is no problem of non-equivalence (at
word level) between the source and target languages, as the TL possesses a ready
translation equivalent of the SL term in question. Such cases are designated with
the code RTE (ready translation equivalent).
• The code BT in Table 2 above, stands for Back-translation.
The data from Table 2 is quantified in Table 3. (Note that in cases where there are two
translation equivalents for a particular keyword, each of these equivalents is counted as
a half).
TAMA 2003 South Africa: CONFERENCE PROCEEDINGS 86
Table 3: Quantitative analysis of 20 SL items translated into isiZulu and Sepedi
Translation strategy isiZulu Sepedi
# terms % terms # terms % terms
Paraphrase:
• PAR
• PAU
2
3
10
15
3
1
15
5
Term formation strategies:
• LWE
• LWT
• COM
• SSP
1.5
—
2
1
7.5
—
10
5
—
3
2
0.5
—
15
10
2.5
More general and/or neutral word:
• MGW
• MNW
5.5
1.5
27.5
7.5
8.5
—
42.5
—
RTE 3.5 17.5 1 5
CST — — 1 5
Total 20 100 20 100
The following conclusions can be drawn from Table 3:
• In both isiZulu and Sepedi, translation by a more general and/or neutral word
seems to be the preferred strategy, i.e. in a little more than a third of all cases (35%)
in the isiZulu sample and approaching half of the cases (42.5%) in the Sepedi
sample.
• The next most popular translation strategy in isiZulu would seem to be translation
by paraphrase at 25% of the sample.
• This contrasts with Sepedi where term formation is utilised in just over a quarter of
the cases (27.5%) as the next most popular translation strategy after translation by a
more general word.
• Term formation as translation strategy is found in just less than a quarter of cases in
the isiZulu sample (22.5%).
• In Sepedi, translation by paraphrase accounts for another fifth of the sample (20%).
• In only 17.5% of the cases does isiZulu make use of a ready / direct translation
equivalent, whilst in Sepedi the remaining 10% of the cases consists of one
instance of translation through the use of a ready / direct equivalent, and one
instance of translation through cultural substitution.
• The same translation strategy is used in both isiZulu and Sepedi in the translation
of the following SL terms: assessment criteria, definitions, finance / financial,
gender, global, guidelines and redress.
• In a few cases, both isiZulu and Sepedi display synonym richness. This is the case
with the SL terms census, South African Qualification(s) Authority and Standards
Generating Body in isiZulu, and apartheid and research in Sepedi.
R. Gauton, E. Taljard & G-M de Schryver — Translating Terminology into all SAn Languages 87
5. Conclusion
In this paper we have shown how electronic machine-readable corpora can be used in
determining the strategies used by professional translators in finding translation
equivalents for SL terms. This is a wide-ranging project that will require the
participation of researchers from all of the South African languages, and which will
on completion provide a wealth of data with numerous practical applications. Apart
from the obvious benefits of this undertaking for the fields of translation studies and
terminology, the results from this project will provide guidelines to especially
African-language translators, confronted with the onerous task of finding translation
equivalents for SL terms foreign to these languages.
References
Baker, M. 1992. In other words: a coursebook on translation. London: Routledge.
Cluver, A.D. de V. 1989. A manual of terminography. Pretoria: Human Sciences
Research Council.
De Schryver, G.-M. 2002. Web for/as Corpus: A Perspective for the African
Languages. Nordic Journal of African Studies 11/2: 266-282.
Mtintsilana, P.N. and R. Morris. 1988. Terminography in African languages in
South Africa. South African Journal of African Languages 8/4: 109-113.
Prinsloo, D.J. and G.-M. de Schryver. 2002. Towards an 11 x 11 Array for the
Degree of Conjunctivism / Disjunctivism of the South African Languages.
Nordic Journal of African Studies 11/2: 249-265.
Taljard, E. and G.-M. de Schryver. 2002. Semi-Automatic Term Extraction for the
African Languages, with special reference to Northern Sotho. Lexikos 12: 44-74.
Uzar, R. and J. Walinski. 2000. A comparability toolkit: Some practical issues for
terminology extraction. In B. Lewandowska-Tomaszczyk and P.J. Melia (eds.).
PALC’99: Practical Applications in Language Corpora. Papers from the
International Conference at the University of Lodz, 15-18 April 1999: 445–457.
(Lodz Studies in Language 1.) Frankfurt am Main: Peter Lang.
TAMA 2003 South Africa: CONFERENCE PROCEEDINGS 88
