Lexicographic Treatment of Negation in Sepedi Paper Dictionaries

Negation in Sepedi is a complex matter — a variety of negative morphemes are used across complicated verbal and copulative structures and these morphemes are not interchangeable. Dictionary users in a text production situation need to be guided to the correct compilation of negative constructions. In most cases negative morphemes are not lemmatised in Sepedi paper dictionaries or if they are, the lexicographic treatment is inadequate. This paper provides an inventory of negation from which an abstraction of the negative morphemes is given. This is followed by a critical evaluation of Sepedi dictionaries and suggested model entries for lexicographic treatment of negative morphemes.


Introduction
For negation in isiZulu, Bosch and Faaβ (2018: 1) state that "as an important instrument of language use, one would therefore expect aspects of negation to be dealt with in dictionaries". They also emphasize that corpus studies reveal an "impressive number" of complex morphological constructions expressing negatives. The situation for Sepedi is no different. Corpus studies reveal that negative morphemes in Sepedi are frequently used. They are, however, often not lemmatised and treated in their own right in dictionaries. Prinsloo and Gouws (1996) attempted to reduce the negation strategies into a single socalled ga/sa/se convention. This convention has since been implemented in several dictionaries, e.g. Oxford Bilingual School Dictionary (ONSD) and Popular Northern Sotho Dictionary (POP). It is a powerful convention and reasonably user-friendly. The convention given in (1) caters for multiple negative constructions in the indicative, situative, relative, subjunctive, habitual, consecutive, infinitive and imperative moods and for the lemmatisation of verbs ending in -e. With a few exceptions verbs ending in -e are not lemmatised in Sepedi dictionaries.
(1) rêke, rekê must buy; ..ga/sa/se..~ not buy What is required, however, for negative morphemes is that they should be separately lemmatised with full treatment. Traditional dictionaries reflect a strong word-bias, not focusing on items smaller than words such as morphemes, or multiword items, thus literally taking the alternative term for lemma, i.e. headword as the norm. This issue, however, falls beyond the scope of this article, cf. Gouws (1989) and Gouws and Prinsloo (2005) for a detailed discussion.
The aims of this article are to (a) give an overview of negation in Sepedi, presented as example driven rules, (b) extract the negation rules, (c) determine the frequency of use of negative morphemes in a Sepedi corpus, (d) critically evaluate the treatment of negation in Sepedi dictionaries and (e) compile model entries for Sepedi paper dictionaries. These model entries will firstly incorporate all the negative morphemes that have to be lemmatised and secondly, give a detailed scope of all the types of information and lexicographic conventions that have to be considered in the treatment of these lemmas.

Overview of negation in Sepedi
Negation is discussed in detail in Sepedi grammars such as Lombard, Van Wyk and Mokgokong (1985) and Poulos and Louwrens (1994) but is fragmented over different sections for the different verbal moods and copulatives. Inexperienced learners could find it difficult to get a full perspective on negation from grammar books. This article departs from such typical grammar-book descriptions but presents a summary of the full scope of negative constructions in only three tables. These tables will reflect an example-driven approach in a systematic way across different verbal moods and copulatives. The tables will also be used as a basis for the extraction of the negation rules, compilation of text boxes, running footers, back matter components, etc. in paper dictionaries. For the interested reader the full scope of verbal moods and copulatives is summarised in appendices A and B.

Negation of verbal moods
In this article the distinction of eight verbal moods for Sepedi by Lombard et al. (1985) is followed. The negation rules indicated in tables 1 to 3 are intended for a step-by-step build-up of the negative morphemes which will be regarded as lemma candidates. These negative morphemes will also be used in the evaluation of 12 Sepedi dictionaries. Tables 1 to 3, as well as appendices A and B, can also be used in the back matter of dictionaries as a reference address for the articles of negation morphemes in the central text. So, for example, references from the articles for ga, sa and se 'not' in the central text to the back matter contextualise ga, sa and se in the broader perspective of verbs and copulatives and could be valuable for the user's cognitive needs.

4.
Negation of copulatives Louwrens (1994) gives a brief description of copulatives as follows: A term used to refer to structures which incorporate a variety of prefixes (called particles in some grammatical descriptions of Northern Sotho) of which some resemble the subject concord (e.g. o bohlale, le bohlale, e bohlale, etc. 'he/it is clever'), whereas others have a unique form like ke (e.g. ke monna 'it is a man') and ga se (e.g. ga se monna 'it is not a man'), as well as a series of verb forms with the copulative verb stems -bê (e.g. … gore a bê bohlale 'so that he may become clever'), -ba (e.g. … a ba bohlale 'and then he became clever'), -le (e.g. … gê a le bohlale 'if he is clever'), -se (e.g. … gê a se bohlale 'if he is not clever') and -na (e.g. o na le lehufa 'she is jealous'). The copulative expresses the English verb to be, and is therefore translated with 'is', 'was', 'will be', 'has been ', etc. (Louwrens 1994: 40) The copulative in Sepedi is a very complex system -see appendix B for a detailed overview of the structure of static and dynamic copulatives. As for the verbal moods in table 1 the aim of the following section is to list the different negative copulative constructions and to extract the negative morphemes which will also be regarded as lemma candidates for Sepedi dictionaries.

Static copulatives
Louwrens (1991: 71) defines static as "referring to a state of rest which is characterised by the absence of motion or change". Three semantic relations are distinguished between a subject and a complement within the copulative, i.e. identification/equality, descriptive or associative.

Dynamic copulatives
Louwrens (1991: 71) defines dynamic as "referring to a state in which things are in motion and therefore changing". Dynamic copulative constructions are therefore often translated as '(to) become'. Similar to static copulatives, dynamic copulatives distinguish identifying, descriptive and associative relations between the subject and complement.

Identifying dynamic copulatives
As in the case of identifying static copulatives in table 2, first and second persons use their own subject concords, i.e. ke, re, o and le but third persons use a neutral subject concord e. In table 3 only examples with third person class 1 monna 'man' as subject will be given.

Descriptive dynamic copulatives
Descriptive dynamic copulatives use the same negation strategies as the descriptive identifying copulatives, cf.

Associative dynamic copulatives
Associative dynamic copulatives use the copulative verb stem ba instead of na which is used in the associative static copulatives (cf. 4.1.3, table 2). Associative dynamic copulatives occur in all of the moods and follow the same negation rules as ordinary verbs as in table 1. Associative dynamic copulatives also use the same negation strategies as associative static copulatives except for the situative and relative where the negation morpheme -sa is used in associative dynamic copulatives, e.g. a sa be le 'not be with' instead of -se (4.1.3.2, table 2) for static associative copulatives. Due to space limitations the associative dynamic copulatives will also not be presented here as a separate full table for the different moods.
The reference numbers given in the first columns of tables 1 to 3 for the suggested lemma candidates ga, sa, se(go/ng), ga se and ka se in the final columns of these tables, are intended to guide the lexicographer in the microstructural treatment of these lemmas. The lexicographer should, namely, try to include examples representing the moods, morphological changes in subject concords, verbal endings, grammatical structures, etc. in following the reference numbers for each negation morpheme. See discussion below on the suggested scope for treatment and utilization of the reference numbers.

Frequency of use of negative morphemes in a Sepedi corpus
Negative morphemes are frequently used in Sepedi. Consider the following approximate counts for ga, sa, se, ga se and ka se in a corpus of 1.2 million tokens in table 4. 1

A critical evaluation of the treatment of negation in Sepedi dictionaries
Twelve Sepedi dictionaries were studied for their lemmatisation and treatment of the negative morphemes ga, sa, se, ga se and ka se and will be briefly evaluated in this section.  (SEPD) and Woordeboek Afrikaans-Noord-Sotho (WANS). Full references are given in the bibliography. First, consider table 5 for ten dictionaries where Sepedi is the target language in English-Sepedi or, Afrikaans-Sepedi dictionaries. In English we often shorten not to n't. For example, You didn't close the door. In Northern Sotho, the notion of not (doing something) is expressed by a variety of negative morphemes, often combined with a change in the verbal ending. Examples of negative morphemes are shown as translation equivalents at the start of this dictionary entry.

not be verb (isn't, not being, wasn't/weren't not been) See ISN'T, AREN'T
The verb not be is very irregular. In the : Dynamic copulative constructions are therefore often translated as '(to) become)'present tense, for example, we have: I am not, you aren't/are not, she/he/it isn't/is not, we aren't/are not, both/all of you aren't/are not, they aren't/are not. ONSD also provides frequency and part of speech information as well as additional information in text boxes.

SEPD
In tables 6 to 8, author inserted English translations for Afrikaans are given in double square brackets. There is no entry for ga in PTLH but entries are given in the other nine dictionaries. SEPD has an entry for ga but does not lemmatise the possessive, place of and situative particle if, when as separate entries. This runs against the lemmatisation policy followed in this dictionary i.e. to separate such homonyms and to allocate homonym numbers. POP also treats the negative morpheme ga and the locative ga as well as if, when in the same entry. This is also the case in PUKU where nie [not] is given amidst four other unrelated translations. At least one of the examples given illustrates the use of ga in an identifying copulative containing the progressive sa.  1.3(1), table 1) and indicative descriptive copulative respectively. The examples also indicate two other important aspects, i.e. the use of subject concords of the first person as well as class 1 and that the past tense form of the verb changes back to the present tense. They also illustrate that unlike English, for example, Sepedi does not distinguish between he is, she is, they are, and it is -they all convert to ke 'it is' and ga se 'it is not' in the negative. SEDIBA also gives exam-ples of ga se for both verbs and copulatives. PUKU'89 indicates tone by means of the letters L 'low' and H 'high'. GNSW also indicates tone e.g. high tones é and ό -low tones are unmarked. ONSD's entry for ga is of high lexicographic quality. It gives detailed grammatical guidance stating that ga is a morpheme, used in the negative of the indicative and that the verbal ending changes to -e illustrated by examples. It also gives frequency information by means of a one, two or three star convention -indicating in this case that ga is one of the top most frequently used tokens in a Sepedi corpus, cf. table 4. In table 7 no entries were given for sa in SEPD, NAAN and PTLH. NEN and PUKU treat the negative morpheme sa within the article of still. If the user reads through all four entries given for sa, (s)he will eventually find the information given for the use of sa in negative constructions but the possibility of missing the information on negation is high. POP does even worse by treating the possessive, negative, subject concord and progressive in a single entry for sa. GNSW does well in stating that sa is used in negation of the infinitive and situative moods, illustrated by examples. SEDIBA gives the relevant basic information with an example from the situative. ONSD does well in providing a homonym number, frequency indication, part of speech as well as morphological information and a translated example in a relatively short entry. PUKU'89 is the only dictionary indicating negation with sa in the relative mood.  Lombard et al. (1985: 188) regard ke as an auxiliary verb stem. From a user perspective se ke should be lemmatised and treated or at least be treated in the articles of se and ke. SEPD has an entry for se ke with translation equivalents do not, don't. GNSW treats se ke in the article of ke with reference to its seventh lemma for ka. POP has an entry for se ke with translation equivalent must not but under the lemma seke. is not lemmatised in any of the dictionaries but ONSD gives ka se as part of an example for sa in table 7. A model entry for ka se will be suggested in the next paragraph.

The compilation of model entries for paper Sepedi dictionaries for the extracted negative morphemes
In the final columns of tables 1 to 3 above the negative morphemes ga, sa, se(go/ng), ga se and ka se were extracted from the negation rules for verbal moods and copulatives. These morphemes are proposed as required lemmas in the macrostructure in Sepedi dictionaries. It will also be attempted to formulate the required essential scope of information types and lexicographic conventions to be considered by the lexicographer in their microstructural treatment in terms of grammatical and semantic information. The lexicographer must strive to cover the full scope of use of a specific negation morpheme in the lexicographic treatment. So, for example, should ga se not only be lemmatised but the compiler should also cover its use in verbal moods as past tense negative (3.1.3(1)(2), table 1) and as negative copulative (4. 1.1.1, table 2). Such guidance is essential for especially text production with copulatives. No formal user studies of negation has as yet been done for Sepedi but a common error observed in tests of undergraduate students is incorrect negation in the indica-tive of the static, identifying copulative (4. 2.1.1.1, table 2). Learners would for example correctly construct the copulative I am not a teacher as ga ke morutiši but incorrectly apply negation with ga to the man is not a teacher as *monna ga o morutiši or even *monna ga ke morutiši instead of monna ga se morutiši. From the 10 dictionaries consulted for their treatment of ga in table 6, only ONSD provides the required guidance by lemmatising and treating ga se 1 . Lexicographers should firstly decide on the generic scope of information types to be presented, i.e. which data types should be considered in all cases. They should then decide on the conventions, structural markers and layout strategies to be used consistently. Consider the following recommended information types for negative morphemes in Sepedi: -Homonym differentiation, preferably by homonym numbers presented as superscripts, cf. ONSD, SEPD, PUKU'89 in tables 6 to 9. -Frequency information by means of a specific convention such as stars or filled/unfilled diamonds, cf. ONSD in tables 6 to 9. -Part of speech.
-Grammatical and text production guidance, cf. ONSD in tables 6 to 9. Note, however, that there are no binding rules for conventions and structural markers but the lexicographer will be well-advised to follow the practices that users became used to in other dictionaries such as presenting the lemma in lowercase and in boldface, translation equivalents in Roman, etc.
The suggested scopes and model entries for each of the negative morphemes ga, sa, se, ga se, and ka se are briefly presented in the following paragraphs.
(2) ga 1 *** negative morpheme ■ (generally marks the negative form of indicative verbs) ♦ ga ke nyake go robala ka pela. I don't want to go to sleep early. ◊ ga (...) [+ VERB ENDING in -e] ■ do/does not ♦ Bašemane ga ba rate go bala. Boys do not like to study.
A reference to the lemmas ga se 1 and ga se 2 is advisable so that the user will not miss the treatment of the indicative past for verbs (3.1.3(3)(4), . This entry could be further enhanced by an example from class 1 to indicate that the subject concord changes from an o in the positive to a in the negative. Indication of circumflexes as in POP and PUKU is also recommended.
For sa the scope includes situative present ( SEDIBA provides good entries for se given in (4): (4) se nie not -ye gae moenie huis toe gaan nie don't go home ka -mag/kan/sal -may/ can/shall/will -ankê o -tlê moenie kom nie (jy) don't come (you, s.) ga -ka ja ek het nie geeet nie I did not eat mêêtse a -kê a bela die water moet hoegenaamd nie kook nie the water should not boil whatsoever -sa nie meer nie no longer se nie w. (negatiewe kop. ww. b. not (negative cop. v.) re-kgolê terwyl ons nie ver is nie while we are not far away e bê e-hy/sy was nie he/she was not SEDIBA's entries are strong on coverage, they give examples from the imperative, future tense, subjunctive mood, indicative past tense, use with the auxiliary verb stem -ke, negation with the progressive, copulative descriptive, and even some guidance on the use of a copula in a past tense construction. The run-on layout, although indented, is less user friendly, e.g. compared to ONSD's entry for se which also gives grammatical and frequency information.
Homonym indication by means of superscript numbers is recommended.
(5) ga se 1 *** copulative particle ■ is/are not ♦ Batho bao ga se ba lapa la ka. Those people are not my family. ga se 2 *** negative morpheme ■ did not (marks the negative form of past tense indicative verbs) ♦ Ga se ke bone gore ke sefatanaga sa mang se se mo fološitšego mo kgorwaneng. I did not see whose car dropped her off at the gate.
Consider also the suggestion for model entries for ga se by Prinsloo (2002: 36) in example (6): (6) ga se [cop. part. neg.] it is not, ga se phošo ya gago it is not your fault; he/she/it is not, Satsope ga se morutiši, ke mongwaledi Satsope is not a teacher, she is a secretary; they are not, dingaka ga se mahodu doctors are not thieves • ke • BM1.1; ■■■■■ … se, ga ~ • ga se In (6) a cross-reference marker "•" informs the user that more information is available in the back matter of the dictionary. For the user who looked up se but actually need information about ga se a reference entry is given at se with a cross-reference to ga se where a full treatment is given.
Provision of a text box at ga se 1 in (5) for indicative, identifying, static copulatives (4.1.1.1, table 2) is recommended as learners or inexperienced users are often unfamiliar with the fact that no distinction is made in Sepedi for he is not, she is not, they are not and it is not and often result in incorrect utterances for, e.g. he/she is a teacher or they are teachers such as *ga a morutiši or *ga ba barutiši respectively.
Treatment of ga se 2 should include examples from the alternative negative strategies (3.1.3(1)(2), table 1) and could also be nicely complemented by a page footer containing the set of alternative concords used or a reference to the front, middle or back matter where all the alternative concords are presented together.
As mentioned above, none of the dictionaries lemmatised ka se. Consider an attempt towards a model entry for ka se in (7).
(7) ka se 1 *** [kɑ se] (negative of the potential ka) ■ can not, may not. Maripane a ka se kgônê. Maripane will not be able to nka se arabê potšišô yê I can not answer this question ka se 2 *** [kɑ se] (negative of the future tense indicator tlo) ■ shall/will not re ka se boê we shall not come back gê ba ka se arabê if they will not answer yô a ka se ithutêgo ka mafolofolo he who will not study hard Provision of a text box at ka se is also recommended to inform the user about an important difference between the negative of the future in English and Sepedi.
1 st Person singular ka, 1 st Person plural ra, 2 nd Person singular wa, 2 nd Person plural la, Class: 1 a, 2 ba, 3 wa, 4 ya, 5 la, 6 a, 7, sa, 8 tša, 9 ya, 10 tša, 14 bja, 15 -18 gwa Unlike in English there are not different constructions in Sepedi for he is not, she is not, they are not and it is not -all convert to the same construction: ga se 'it is not' The examples in ka se 1 indicate that the subject concord of class 1 is a and that ke + ka changes to nka and in ka se 2 illustrate and confirm the use of ka se in the indicative, situative and relative. In this way the use of examples are not merely intended to confirm the existence of a word in the language but adds to the meaning information.
The text boxes at ga se and ka se appropriately illustrate the nature and purpose of text boxes in terms of Gouws and Prinsloo (2010: 501): Text boxes are salient dictionary entries and as such they are used to place more than the default focus on a specific data item … guidance in terms of sense, contrasting related words, restrictions on the range of application, register, pronunciation, et cetera. … Where dictionaries have a text production function data could be included in a text box to emphasise the use or non-use of certain combinations and collocations as well as proscriptive guidance.
As far as the required information for the treatment of negative morphemes is concerned, one should not attempt to compile a one-size-fits-all set of information types. For paper dictionaries the lexicographer should keep in mind that the size of the dictionary, i.e. available physical space for an article is always a crucial consideration. Therefore, one cannot expect lengthy detailed articles for negative morphemes in small bilingual dictionaries such as POP, NEN, PUKU, PUKU'89, SEDIBA, etc. The challenge is maximum utilization of limited space and from the dictionaries discussed in reference to tables 6 to 9, ONSD comes closest to covering the most important information types in relatively short articles and in a user-friendly layout.
As a final suggestion for model entries consider also the front matter entry in figure 1: In Sepedi there is no distinction between the negative forms for shall/will not (future tense) and can not (potential), both are expressed by ka se. You have to determine from the context which meaning applies.  figure 1 in the user's guide to the dictionary will guide the user to appreciation of the information types presented in the articles.

Conclusion
Within the space limitations of an article it has been attempted to give an overview of negation in verbs and copulatives and the frequency of occurrence of negative morphemes in Sepedi. The focus was on the different negation strategies and the compilation of a list of lemma candidates to be treated in dictionaries for this language. A critical evaluation of a number of Sepedi dictionaries and suggestions for the scope and conventions for the treatment of negative morphemes were given. Model entries for dictionary articles for negative morphemes and user guidance in the front matter as well as comprehensive back matter entries were proposed.