How to deal with so-called "vowel verbs with variant non-vowel forms" in a bilingual Zulu lexicon ? The case for a pragmatic approach

Abstract

Whilst lemmatisation in Zulu (and cognate languages) remains to this day a partially unsettled issue, with dictionaries, both bilingual and monolingual, adopting differing strategies in respect of nouns, we wish to focus here on a somewhat minute aspect of the matter: how to lemmatize verbs with optional initial vowels. As the variation occurs at the initial, it is crucial that it receives proper dictionary treatment, lest the user be misdirected or misinformed.

Full text

Journal of the Digital Humanities Association of Southern Africa, Vol. 4, No.1
How to deal with so-
called "vowel verbs
with variant non-vowel
forms" in a bilingual
Zulu lexicon? The case
for a pragmatic
approach.
Michel Lafoni
research fellow (retired),
CNRS Llacan Paris, IFAS - Johannesburg
& CenterPoL, University of Pretoria,
maikoro12@gmail.com
& Bolofo Mongezi
PhD Candidate, Wits School of Education
Abstract
Whilst lemmatisation in Zulu (and
cognate languages) remains to this day a
partially unsettled issue, with dictionaries,
both bilingual and monolingual, adopting
differing strategies in respect of nouns, we
wish to focus here on a somewhat minute
aspect of the matter: how to lemmatize verbs
with optional initial vowels. As the variation
occurs at the initial, it is crucial that it
receives proper dictionary treatment, lest the
user be misdirected or misinformed.
Keywords: isiZulu - lexicography - vowel
verbs
We shall start by presenting verbs with
optional initial vowels in Zulu, contrasting
the lemmatization strategies offered by main
available dictionnaries, some of which in
our view lack consistency. We shall then
introduce to the pragmatic solution we
adopted in our bi-directional French to Zulu
and Zulu to French lexicon, which we claim
gives better justice to the language.
1 The corpus
Zulu (or isiZulu)ii is in terms of
demography the main African (Bantu)
language spoken in South-Africa,iii being in
its different sociolects the first or home
language of close to 13 millions people, viz.,
22% of a population of at least 59 millions.iv
Although there exists a number of regional
or sub-regional dialects, its standard form,
taught in schools and propagated through
formal situations, including literature, the
written and spoken press, as well as official
ceremonies, is widely accepted. It also
boasts a well-accepted orthography. This
paper is based on the standard language and
our data is taken mainly from secondary
sources, id est, dictionaries as quoted in the
paper, complemented by random
observations taken from formal and
informal written documents as well as
informal speech, and the native-speaker
fluency of one co-author.
Zulu has a rather large inventory of
what (Clement Martyn Doke 1992, 131)
refers to as "vowel verbs with variant non-
vowel forms" , such as
-edlula or -dlula,v passvi
-esutha or -sutha, be satiated
with food
The initial vowel is mostly /e/, but /a/
and /o/ also occur, more often as further
variants (examples from dictionaries):
1

Proceedings of the 3rd Workshop on Resources for African Indigenous Languages 2022
-ehlula, -ahlula or -hlula,
conquer, defeat
-ejwayela, -ojwayela or
jwayela, be used to
-ephula, -aphula, -ophula, or
phula, break
Whereas the presence in the standard of
variant forms may proceed from the
incorporation of regional features as,
although based mainly on «the Central
Zululand Dialect» (Kubeka 1979, 83), it
includes arguably features of other varieties,
it seems that the variation has now become
evolutionary, as some of Doke’s vowel-
initial verbs do not seem valid today (but
see further down):
?-efunda for -funda study
?-egcwala for -gcwala, be full
?-emithi for -mithi be with small
(animals); become pregnant
?-ojwayela for -jwayela, be used
to
This would point to a phenomenon of
vowel erosion. However the fact that some
of those verbs, like -gcwala, seem to have
their source in ideophones (gcwa, of being
full to the brim) which are consonant-
commencing suggests this explanation is not
sufficient. The vowel would then have been
added (or the ideophone would have been
obtained by deletion of the vowel). We shall
not investigate this issue further however, as
it is not relevant to the problem at hand.
Nouns derived from the same root
show the same alternative:
abalusi or abelusi, herders (cf.
elusa or -lusa, herd)
but all possibilities are not always
admitted:
umelaphivii healer: ? umlaphi,
The same variation may apply to
extended verbs and nouns derived from
them:
-ehluleka or -ahluleka, fail and
isehluleki or isahluleki failure
-efundisa or -fundisa teach, and
abafundisi or abefundisi, priests
-efundisa seems clearly
obsolete but the noun
abefundisi is attested
It also happens that a variant ∂ does not
cover all meanings of variant ß, suggesting
that the verbs were originally different but
became conflated due to the phonetic loss of
the initial vowel:
-phuza has now two meanings:
phuza, drink and -phuza
(variant of -ephuza) be late
To this category we add the three (or
four) verbs with /i/ as the initial vowel, that
Doke in the passage quoted above refers to
as “latent-vowel verbs” due to the fact that /i/
may only appear in conjugated forms when
preceded by a vowel /a/with which it then
coalesces resulting in/e/:viii
-(i)ma: stand: ngiyema or
ngiyama, I stand
-(i)-za, come: bayeza or bayaza,
they are coming
-(i)zwa, understand: ubezwa or
ubazwa, listen to them
Since in Zulu /i/ tends to disappear
when following any vowel other than /a/
(or merge if the preceding vowel is /i/), no
conclusion can be drawn from instances
2

Journal of the Digital Humanities Association of Southern Africa, Vol. 4, No.1
where initial /i/ does not surface, as to what
is the variant of the verb stem in any given
instance:
ngimile, I stood: |ngi|-|im|-|ile|
or |ngi|-|m|-|ile|
uzwile, you heard: |u|-|izw|-|ile|
or |u|-|zw|-|ile|
We therefore posit for those verbs a
variant form of the stem with an initial
vowel, akin to the situation above:
-ima or -ma, stand
-iza or -za, come
-izwa or -zwa, hear, understand
The choice of variant appears related to
the tense form, the presence of an extension
after the root which seems to result in a
preference for the shorter, consonant-
commencing form, as well as to the
stemitself. The following exemples are
drawn from our own observations of
unsollicited speech:
ngiyezwa rather than ngiyazwa,
I hear
ngiyazwisisa rather than
ngiyezwisisa, I understand very
clearly
kuyezwakala or kuyazwakala, it
is understandable
ngiyeza rather than ngiyaza, I
am coming
ngiyabazisa endlini rather than
ngiyabezisa endlini, I make
them come home
ngiyema or ngiyama, I stand
ngiyamela abafazi rather than
ngiyemela abafazi, I am waiting
for the women
2 Treatment in Dictionaries
How are these facts dealt with in major
Zulu dictionaries?
Keeping to items already mentioned,
we compare the main dictionaries presently
available:ix bilingual - Doke & al. Zulu-
English and English-Zulu (Clement M. Doke
et al. 1999), Dent and Nyembezi’s Scholar’s
Zulu (Dent and Nyembezi 1995), de
Schryver’s (de Schryver 2015)- and
monolingual - Nyembezi’s (Nyembezi 1992)
and Mbatha’s (Mbatha 2006). We are aware
that these dictionaries were not compiled
within the same framework. Whereas the
four older follow with small deviations
between them the strategy set out by Doke
for Zulu lexicography, which lemmatizes
stems rather than words, de Schryver’s,
besides being corpus-based, introduces an
approach which seeks to lemmatize words
rather than stems (de Schryver and Wilkes
2008). However, this revolution in Zulu
lemmatization does not really affect verbs.
Even though Doke and Nyembezi quote
verbs under the imperative while other
dictionaries refer to the verb stem,
appropriately defined in (Marlo 2013) as “an
obligatory root, one or more possibly occurring
derivational suffixes [also known as
extensions]x and an inflectional final suffix
commonly called the “Final Vowel”, the form is
segmentally similar, the singular imperative
being none other in Zulu than the said stem
with a specific tone structure.
We use the following symbols:
A: main entry (as revealed by
length of description);
3

Proceedings of the 3rd Workshop on Resources for African Indigenous Languages 2022
A’: main entry although shorter
than A [that implies two main
entries in same dictionary]
A + a: main entry + cross-
reference to at least one variant;
B: shortened entry, with cross-
reference to main;
B’: cross-reference to main;
0: no mentionof the item
-phula -ephula -aphula -esutha -sutha -egcwla -gcwala
break be satiated become full
Doke B B A + a A’ A B A + a
Dent A A A A’ A 0 A
de
Schryver
A 0 0 0 0xi 0 A
Nyembezi A’ A A’ A’ A 0 A
Mbatha A A’ A’ A A 0 A
-dlula -edlula -hlula -ehlula -ahlula -emithaxii -mitha
pass conquer become pregnant
Doke A + a B B B A +a a A + a
Dent A B’ A A A A 0
de Schryver B’ A B’ A B’xiii 0 A
Nyembezi A A’ A A’ A’ A’ A
Mbatha A A’ A A’ A’ 0 A
4
-ima -ma -iza -za -imba -mba -izwa -zwa
stand come dig hear
Doke 0 A + a 0 A + a 0 A + a 0 A + a
Dent 0 A 0 A 0 0 0 A
de
Schryver
0 A A 0 012 A 0
Nyembezi 0 A 0 A 0 A 0 A
Mbatha 0 A 0 A 0 A 0 A

Journal of the Digital Humanities Association of Southern Africa, Vol. 4, No.1
3 Observations
It seems, even judging from such a
limited and haphazard sample, that
inconsistency prevails within dictionaries as
well as between them, as a cursory look at
the table suggests.
Zulu to English.
Since Doke refers specifically to "vowel
verbs with variant non-vowel forms" one would
expect these verbs to be entered
systematically under the vowel with an
indication that a consonant-commencing
variant exists: not so but most instances
include a cross-reference, which mitigates
the issue. In Dent on the other hand
repetition is frequent: -ephula, -aphula and
-phula among other such examples
constitute three different entries with almost
the same definitions repeated without any
cross-referencing. De Schryver is more
consistent in cross-referencing the
(supposedly) less frequent form to the main
one.
Regarding the so-called latent vowel or
i-commencing verbs, Doke is the only one to
systematically make mention of ”latent i” in
the description of all corresponding entries.
Dent and de Schryver make no explicit
allusion to it while implying it nevertheless
through examples: under -zwa, Dent gives
ukungezwa nakutshelwa, to want to see by
oneself (“not wanting to be told”), where the
negative marker -nga- alters to nge as its
/a/ vowel coalesces with the initial /i/;xivxv.
In a similar fashion, under -za de Schryver
gives abantu beza ngobuningi, people came
in their numbers, whereas beza cannot be
obtained from the subject prefix ba- and the
verb given as -za. Same situation under
zwakala (separate entry from -zwa) with
kuyezwakala it is understandable.xvi
English to Zulu
In all three bilingual dictionaries, only
the form given as main entry in the Zulu to
English is indicated, with no mention
whatsoever of variants. That applies inter
alia to Dent where, under break, only -aphula
is listed.
Zulu monolingual
Nyembezi and Mbatha provide almost
similar descriptions for each variant, each
treated as a main entry (see -dlula, -edlula
inter alia), occasionally indicating the
existence of a variant assimilated to a
synonym like in the -hlula and -phula
series. Even if variants are somehow
synonyms, it would make sense in a
linguistic work to discriminate. And that
does not preclude inconsistencies: one of
Mbatha’s examples under dlula is
ukwedlula ngendlu yakhiwa not to offer
assistance to people working, which should
appear under -edlula. Same for uwephulile
umoya wami lo mfana, this boy broke my
heart, which should illustrate ephula rather
than -phula. As for the “latent vowel”, no
explicit mention whatsoever. However
Mbatha includes in his examples conjugated
5

Special Issue 2022: Crossroads Digital Humanities
forms where the “latent i” does appear:
under -za: (…) uyeza, (…) you are coming;
under -zwa: (…) sengathi akezwa (…) as if
he does not understand; ukungezwa
ngokutshelwa (see above and note 14).
4 A pragmatic approach
Our bidirectional French / Zulu lexicon
(Lafon and Mongezi 2022) has no pretention
to be extensive but rather a handy support
whilst one is engaged in conversation. Space
and cost were huge concerns. Still we aimed
at covering as much vocabulary as possible
whilst attempting to remain consistent
throughout.
In the French to Zulu part, we opted to
place the possible vowel within brackets, so
as to offer a maximum of information in a
minimum of space, treating i-commencing
verbs in a similar fashion:
passer (to pass): -(e)dlula;
aller (to come): -(i)za
Thus all variants of conjugated forms
can be deducted from the lemmas as they
stand.
Obviously the same strategy could not
be followed in the Zulu to French part. In
order to avoid repetition as well as
omission, we opted to lemmatize the form
which appeared to us the more common
(based on native speaker’s intuition), the
other or possibly other form(s) being
however systematically listed and cross-
referred to the one chosen as lemma:
-edlula: voir [see] -dlula
-iza: voir [see] -za
This strategy counterbalanced at least
partially any wrong estimate on our side of
the relative frequency of variants. The
strategy obviously is not ignored by all of
the dictionaries reviewed;xvii rather it is not
applied systematically.
5 Conclusion
It would seem, from the review of
lexicographical treatment of a few items in
old as well as recent Zulu dictionaries, both
bilingual and monolingual, traditional as
well as recent and corpus- and frequency-
based (de Schryver’s), that the twin issues of
"vowel verbs with variant non-vowel forms" and
“latent-vowel verbs” have been largely
overlooked. Minor as this may seem, it
remains a blind spot. Our ad hoc solution, to
consistently include all variants in the
simplest and less cumbersome manner,
might then go some way towards achieving
better coverage in that particular instance.
This reminds us that there is no Holy Grail
to achieving consistency and user-
friendliness in the lexicography of Zulu and
cognate languages. Lexicographical issues
have to be considered taking into account
their specifics, this being the only manner to
provide adequate, complete and easily
accessible information.
6 References
         
6

Journal of the Digital Humanities Association of Southern Africa, Vol. 4, No.1
   !    "  
#$  %&    $  
'    (  )  *+
,      -.    /&
  .!(      0&  
 1((  '/ 
  "&  (   )    (
)2.$30
'  4  (  56   
-"(2.
1#'*0.!(.2
07    +  
  869  "(2.
00
:'());)(4)
0::        -:2  
    
859 <$"
:'()= >.:
1((  8=?9    $  '&  @
*(0
/.  0(:    = ,2(2@
0!.  -    /  '  "
".        *
#,A**BC
;.:  ,  0.  ?  D "(
0$          !  
7  <$    !  &    
3
*  )      )2  ===
*>E  FG+  ,((
0  ;  ;@F  )
;(
)  )  #  =H  D-.      
*  )&  "  
#)7*B,B
H?I=HJ
).  )  %  =6 ,(2@  0
"(2.0
!(.2  '  *  0.  = K
,2(2@  0(  !(
"(2.#%".
0(  #'  =H    ;  '@
.
0$  1)      = F
,!  *  ,  ,,!
,!,  ,2(2@  0:
0    =    '&  @
%><$"
0$1) :
=5  D<F      
  >    '(&>&7  ,
"    .      
4'  5=?IH6 ,
<$*&&
7

'&
 ,&&&0AC@&>
(& $@$&(
$@(FLMGMN>
 $(((
(&*((.
(2.&>((
$ >&(='@E&&==
(see https://www.statssa.gov.za). &..$O((
===&.
$-.E(&.&O(&(
(&$E&&(0.@
$ $(..(($@
($@(
$ /&&:&Uzalo =?!$(.=50'
$ :ivaincrease.@:&.-iphaikha(($
$-phagivekhadraw water&$..((&@(
(
>  $&$.@:A'56C(Samuelson 1923)A
C.@((:
>%
> ,(&&((.0$@&22(E(@
(ngisuthi,I am satiated
> -emithimithi:
> !&P.&$ahlulwaehlulwa
>$ 'ukungavumi ukutshelwanot accepting to be told @nga(&

>$ 7$.>(&:&$.(&QMQ@

>$ &$@(($
@.$@$.(&@
&+@&&(
>$ 0((..@!.A/.=C