NCHLT: Sepedi POS tag set

Tag set

The following discussion reflects on broader aspects of the tagging process and tag selection strategies as well as aspects of granularity. A separate table of the actual/working tags will be provided separately.

For purposes of annotators, this tag set is by and large taken over from Taljard et al (2008) and various documents compiled by G Faasz and U Heid from the IMS, Stuttgart and D J Prinsloo and E Taljard, University of Pretoria. The information below refers to the current state of the tagset, but further development will probably necessitate any number of changes.

The tagset is mainly based on the lexical and morphological criteria defined by Lombard (1985) and Louwrens (1991). As described above, the logical structure of the tagset is divided into two layers of linguistic description (annotation levels):

The first annotation level includes all mandatory, or, according to EAGLES, obligatory information, namely up to three elements: an element hinting at the word class, a second one specifying functional or syntactic properties, and a third one giving morphological specifics, cf. e.g. PRO(noun)EMP(hatic)PERS(on).

The second level of annotation includes recommended and optional information. This level is in most cases used for a detailed description of closed class items described in the tagger lexicon. Compare the following excerpt:

Figure 1: Annotation levels

Description	Tag 1st level (mandatory information)	Tag 2nd level (optional/ recommended information)
Pronouns:
emphatic personal	PROEMPPERS	1sg,2sg,1pl,2pl
Verbals:	V	tr
Morphemes:
deficient	MORPH	def

As for the actual tagging, an additional first level of tagging is envisaged. On this level, linguistic words will be tagged. For Northern Sotho, this implies that the four orthographic units ke + a + mo + rata will be tagged as V, since together they constitute a linguistic verb.

The tagset currently distinguishes 29 categories and different levels of annotation. The first part of the tag gives a general indication of the nature of the unit in question. These are as follows:

1. $ = Punctuation

2. ABBR = abbreviation

3. ADJ = adjective

4. ADV = adverb

5. ASP = aspectual marker

6. AUX = auxilliary verb

7. CCOP = class-indicating copulative subject concord

8. CD = class-indicating demonstrative

9. CDCOP = class-indicating demonstrative copulative

10. CP = class-indicating nominal prefix

11. CO = class-indicating object concord

12. CPOSS = class-indicating possessive concord

13. CS = class-indicating subject concord

14. ENUM = enumerative

15. IDEO = ideophone

16. INT = interjection

17. JUNC = conjunction

18. MNEG = negative morpheme

19. N = noun

20. NPP = place and brand name

21. NUM = numerative

22. PART = particle

23. PROEMP = emphatic pronoun

24. PROPOSS = possessive pronoun

25. PROQUANT = quantitative pronoun

26. QUE = question word

27. TENSE = tense marker

28. V = verbal

29. VCOP = copulative verb

As we envisage going deeper into morphological analysis, we also plan for the implementation of the following tags:

AS = adjectival stem

CA = class indicating adjectival prefix

NS = noun stem

NSuf = nominal suffix

VEnd = verbal ending

VExt = verbal extension

VR = verb root

1. PUNCTUATION

The tag $ is used for all punctuation marks. These include full stops, commas, colons, semi-colons, quotation marks, hyphens, exclamation marks, brackets, etc.

2. ABBREVIATION

All abbreviations are tagged as ABBR.

3. ADJECTIVE

The following tags are used:

Level 1: ADJ01-14, ADJLOC

Notes:

• The tag ADJ is used to tag linguistic units which consist of a noun class prefix and an adjective stem

Examples:

se segolo ADJ07

mo gobotse ADJLOC

4. ADVERB

The following tags are used:

Level 1: ADV

Level 2: ADV_loc

Notes:

• Currently, no further specification in terms of the features time and manner is indicated, but adverbs of place / indicating locality are tagged on the second level as _loc.
• Nouns belonging to the locative classes are tagged as NLOC, although their function is that of adverb.
• Nouns to which a locative suffix has been affixed are tagged as nouns although their function is that of adverb. See 19 below.

Examples:

ruri ADV_nil

gaModjadji ADV_loc

5. ASPECTUAL MARKER

The following tags are used:

Level 1: ASP

Level 2: ASP_pot, ASP_prog

Note:

• The deficient verbs forms, also called deficient auxiliary verb forms, -fo, -no, -dio, -yo and -tšo are tagged as ASP.
• On the second level of annotation the progressive morpheme sa and the potential morpheme ka are tagged as ASP_prog and ASP_pot respectively.

Examples:

ba fo bolela	ASP_nil
ba sa bolela	ASP_prog
ba ka bolela	ASP_pot

6. AUXILLIARY

The following tag is used:

Level 1: AUX

Notes:

• The tag AUX is used to tag what is generally termed auxilliary verb stems.

Examples:

ba šetše ba fihlile	AUX
o ile bolela bjalo	AUX

7. [CLASS-INDICATING] COPULATIVE SUBJECT CONCORD

The following tags are used:

Level 1: CCOP01-10, CCOP14-15, CCOPLOC, CCOPPERS

Level 2: CCOPPERS_1sg, CCOPPERS_1pl, CCOPPERS_2sg, CCOPPERS_2pl

Notes:

• Noun class information and information on the person number are included in the tag.

Examples:

le nna ke gona	CCOPPERS_1sg
borotho bo gona	CCOP14_nil
re mo toropong	CCOPPERS_1pl

8. [CLASS-INDICATING] DEMONSTRATIVES

The followings tags are used:

CD01-10, CD14-15, CDLOC

Notes:

• Currently, no distinction is made for the different ‘positions’ or ‘distances’ of the demonstratives. Only noun class information is included in the tag.
• The demonstratives of the locative classes are tagged as CDLOC
• The demonstratives of classes 1a and 2b are tagged as CD01 and CD02 respectively.

Examples:

batho ba CD02

selo seo CD07

felo fale CDLOC

9. [CLASS-INDICATING] COPULATIVE DEMONSTRATIVES

The followings tags are used:

Level 1: CDCOP

Level 2: CDCOP_01-10, CDCOP_14-15, CDCOP_loc

Notes:

• Currently, no distinction is made for the different ‘positions’ or ‘distances’ of the copulative demonstratives. Only noun class information is included in the tag.
• The copulative demonstratives of the locative classes are tagged as CDCOP_loc
• The copulative demonstratives of classes 1a and 2b are tagged as CDCOP_01 and CDCOP_02 respectively.

Examples:

šokhwi CDCOP_01

šedi CDCOP_08

šefale CDCOP_loc

10. [CLASS-INDICATING] NOMINAL PREFIX

11. [CLASS-INDICATING] OBJECT CONCORD

The following tags are used:

Level 1: CO01-10, CO14-15, COLOC, COPERS

Level 2: COPERS_1pl, COPERS_2pl, COPERS_2sg

Notes:

• Noun class information is included in the tag. Tags for first and second person distinguish between singular and plural.
• Object concords that are written conjunctively to the verb stem (e.g. object concord 1st person singular, object concord of class 1 mo- preceding a verb stem with initial consonant b-) are not at this stage tagged.
• The object concords of classes 1a and 2b are tagged as CO01 and CO02 respectively.
• Object concords of the locative classes are tagged as COLOC.

Examples:

Ba re thušitše	COPERS_1pl
Re a go nyaka	COPERS_2sg
Ke a a rata	CO06
Ba tlo se reka	CO07

12. [CLASS-INDICATING] POSSESSIVE CONCORD

The following tags are used:

Level 1: CPOSS01-10, 14-15, CPOSSLOC

Notes:

• The possessive concords of classes 1a and 2b are tagged as CPOSS01 and CPOSS02 respectively.

Examples:

bana ba gagwe	CPOSS02
diaparo tša bana	CPOSS08
fase ga tafola	CPOSSLOC

13. [CLASS-INDICATING] SUBJECT CONCORD

The following tags are used:

Level 1: CS01-10, CS14-15, CSLOC, CSINDEF, CSNEUT, CSPERS

Level 2: CSPERS_1sg, CSPERS_1pl, CSPERS_2sg, CSPERS_2pl

Notes:

• No distinction is currently made between the unmarked form of the subject concord, and the co-called ‘consecutive form’ thereof.
• The tag CSINDEF is used to tag the indefinite subject concord go / gwa
• The tag CSNEUT is used to tag the neutral subject concord e / ya, which is inter alia often found in identifying copulatives.
• Subject concords of the locative classes are tagged CSLOC.
• Subject concord of nouns in classes 1a and 2b are tagged as CN01 and CN02 respectively.

Examples:

se fihlile	CS07
ga di ešo tša fihla	CS10
fase go a tonya	CSLOC
go a fiša	CSINDEF
e be e le marega	CSNEUT
o a tshwenya	CSPERS_2sg
ra thoma mošomo	CSPERS_1pl

14. ENUMERATIVE

The following tag is used:

Level 1: ENUM

Note:

• Linguistic units generally termed enumerative stems in grammars are tagged as ENUM

Examples:

polao e šoro	ENUM
mokgwa o šele	ENUM

15. IDEOPHONE

The following tag is used:

Level 1: IDEO

Examples:

perr	IDEO
sehlee	IDEO

16. INTERJECTION

The following tag is used:

Level 1: INT

Level 2: INT_neg

Notes:

• The tag INT_neg is used to tag the negative interjection aowa.

Examples:

thobela	INT_nil
aowa	INT_neg

17. CONJUNCTION

The following tag is used:

Level 1: JUNC

Notes:

• Conjunctions that consist of multi-word units are not at this stage tagged as JUNC, the tokens making up the conjunctions are tagged separately. Thus e.g. le ge ‘although’ would not be tagged as JUNC, but as le [TAG] and ge [TAG].

Examples:

eupša	JUNC
gore	JUNC

18. NEGATIVE MORPHEME

The following tag is used:

Level 1: MNEG

Notes:

• Negative morphemes ga, sa and se are tagged as MNEG
• The negative form of the identifying copula ke, i.e. ga se is tagged as ga MNEG se VCOP_neg

Examples:

ga ba bolele	MNEG
ba sa bolele	MNEG
gore ba se bolele	MNEG

19. NOUN

The following tags are used:

Level 1: N01-10, N01a, N02b, N14, NLOC

Level 2: _aug, _dim, _loc, _name

Notes:

• Although 17 noun classes plus two subclasses (1a and 2b) are distinguished for Northern Sotho, only 14 noun tags are used. All locative classes (16, 17, 18, N- and ga-) are tagged as NLOC.
• Since nouns belonging to the so-called infinitive class 15 (go goroga) consist of two orthographically distinct units, these are tagged separately; no tag for nouns belonging to class 15 (N15) is therefore used.
• Apart from containing information on noun class numbers, tags for nouns also include the features augmentative (_aug), diminutive (_dim), locative (_loc), name (_name) and even a combination of these, e.g. _dim_loc. These appear after the underscore ‘_’ in the tag. If none of these is present, the absence of such a feature is indicated by _nil.
• It is not necessary to include information regarding singular and plural in the tag, since this information is contained in the class number. Generally, nouns with an uneven class number (1, 3, 5, 7, 9) are singular nouns, whereas those belonging to the evenly numbered classes (2, 4, 6, 8, 10) are plural nouns.
• The distinction singular/plural is not relevant for the locative classes.
• Names of people are tagged as N01a / N02b_name

Examples:

mpho	N09_nil
mpho	N01a_name
mphonyana	N09_dim
mphong	N09_loc
taugadi	N09_aug
ngwakwaneng	N03_dim_loc
fase	NLOC
bomojela	N02b_name

20. PLACE AND BRAND NAMES

The following tag is used:

Level 1: NPP

Level 2: NPP_name, NPP_brand

Notes:

• Place names are tagged as NPP_place, brand names as NPP_brand

Examples:

polokwane	NPP_place
coke	NPP_brand

21. NUMERATIVE

The following tag is used:

NUM

Note:

• For all numbers and numerals (23,1998, iii, etc) the tag NUM is used.

22. PARTICLE

The following tags are used:

Level 1: PART

Level 2: PART_cop, PART_agen, PART_hort, PART_loc, PRT_que, PART_temp, PART_ins, PART_con

Notes:

• The category ‘particle’ is a contentious one and is not recognized by all grammarians. The term ‘prefix’ is often used to refer to members of this category.
• The category ‘particle’ is a closed category.
• The tag PART_cop is used for the copulative particle ke ‘it/he/she is, they are’. The negative form of this copula, ga se is tagged as ga MNEG, se VCOP_neg.
• The agentive particle ke found following after passive verbs is tagged as PART_agen.
• The hortative particle a is tagged as PART_hort.
• Five locative particles (PART_loc) are distinguished for Northern Sotho, i.e. ka, mo, go, kua and ga.
• The question particles (PART_que) are a, na, afa, naa.
• The temporal particle (PART_temp) is used to indicate a point in time.
• The instrumental particle (PART_ins) is ka.
• The connective particle le which is typically found following after the copulative verb stem –na is tagged PART_con

Examples:

ke marega	PART_cop
e bonwa ke dimpša	PART_agen
a re bale	PART_hort
ka kua toropong	PART_loc
na ba tlile?	PART_que
ka Mokibelo	PART_temp
ka thipa	PART_ins
go na le kotsi	PART_con

23. EMPHATIC PRONOUN

The following tags are used:

Level 1: PROEMP01-10, PROEMP14-15, PROEMPLOC, PROEMPPERS

Level 2: PROEMPPERS_1sg, PROEMPPERS_1pl, PROEMPPERS_2sg, PROEMPPERS_2pl

Notes:

• Emphatic pronouns are also called absolute pronouns in some grammars.
• Even when these pronouns appear in collocation with their corresponding nouns, they are tagged as PROEMP, even though in such cases they have strictly speaking, no pronominal function.

Examples:

yena	PROEMP01
rena	PROEMPPERS_1pl
gona	PROEMPLOC
dipuku tšona	PROEMP10
ka yona	PROEMP09

24. POSSESSIVE PRONOUNS

The following tags are used:

Level 1: PROPOSS01-10, PROPOSS14-15, PROPOSSLOC, PROPOSSPERS

Level 2: PROPOSSPERS_1sg, PROPOSSPERS_1pl, PROPOSSPERS_2sg, PROPOSSPERS_2pl

Notes:

• Even though the possessive pronouns of the noun classes of Northern Sotho are morphologically identical to the emphatic pronouns, their status as possessive pronouns is signalled by the possessive concord preceding them.
• Although the status of forms such as gaborena, gabobona and gabolena is uncertain, these forms are tagged as PROPOSSPERS_1pl, PROPOSS02 and PROPOSSPERS_2pl respectively, since they are almost always preceded by a possessive concord.

Examples:

bana ba gagwe	PROPOSS01
bana ba gešo	PROPOSSPERS_1pl
bana ba rena	PROPOSSPERS_1pl
maoto a tšona	PROPOSS10
dikolo tša gona	PROPOSSLOC

25. QUANTITATIVE PRONOUNS

The following tags are used:

PROQUANT01 – 10, PROQUANT14-15, PROQUANTLOC

Notes:

• Even when these pronouns appear in collocation with their corresponding nouns, they are tagged as PROQUANT, even though in such cases they have strictly speaking, no pronominal function.

Examples:

bana bohle	PROQUANT02
tšohle di fedile	PROQUANT10
rena bohle	PROQUANT02

26. QUESTION WORDS

The following tags are used:

Level 1: QUE

Level 2: QUE_N01a, QUE_N02b, QUE_loc, QUE_time, QUE_man, QUE_01 – 10, 14 – 15

Notes:

• In some grammars, the items bjang, neng and kae are described as adverbs, but here they are tagged as question words, asking about manner (_man), time (_time) and place (_loc) (locality) respectively.
• Since the question words mang and bomang can generate agreement of classes 1a and 2b respectively, they are tagged as QUE_N01a and QUE_N02b respectively.
• Information on class numbers is included in the tag for all forms with –fe.
• Note that when –kae is used with the meaning of ‘how many / much’ it appears in an adjectival construction and is consequently tagged as ADJ. Only when it is used with the meaning of ‘where’ is it tagged as QUE_loc.
• Note that when –bjang is used with the meaning of ‘what kind of’ it appears in an adjectival construction and is consequently tagged as ADJ. Only when it is used with the meaning of ‘how’ is it tagged as QUE_man.

Examples:

ba fihlile neng?	QUE_time
ba dula kae?	QUE_loc
batho bafe	QUE_02
o nyaka mang?	QUE_N01a
o rekile eng?	QUE_nil

27. TENSE MARKER

The following tags are used:

Level 1: TENSE

Level 2: TENSE_fut, TENSE_pres, TENSE_past

Notes:

• The future tense morphemes –tlo/-tla are tagged as TENSE_fut.
• When the morpheme ka appears together with the negative morpheme se it functions as a marker of the future tense, and in such cases is marked as TENSE_fut.
• The past tense morpheme –a only occurs in a specific negative form of the so-called past tense and is the only past tense morpheme in Northern Sotho. It is tagged as TENSE_past.
• The so-called present tense morpheme –a which appears only in (some) present tense verbs is marked as TENSE_pres.

Examples:

ba tlo bolela	TENSE_fut
ba a bolela	TENSE_pres
ba ka se bolele	TENSE_fut
ga ba a bolela	TENSE_neg

28. VERBAL

The following tag is used:

Level 1: V

Notes:

• The tag V is used to tag what is generally termed verb stems.
• Prefixes (reflexive prefix, object concord first person singular, object concord class 1 prefixed to verbs stems with initial consonant b-, stabilizing prefix e- with monosyllabic stems) and suffixes that are written conjuntively to the verb root or stem are not at this stage tagged separately.
• The degree of transitivity of verbs is indicated as _tr (transitive), _itr (intransitive), _dtr (double transitive) on level 2.
• Categories such as tense, mood and aspect are not indicated, since these features cannot be assigned automatically by computational means.

Examples:

mmotša	V_tr
ithuta	V_tr
ntshwenya	V_tr
direla	V_dtr
eja	V_tr

29. COPULATIVE VERB

The following tag is used:

Level 1: VCOP

Level 2: VCOP_neg

Notes:

• The tag VCOP is used to tag copulative verbs stems, e.g. –le, -se, -(e)na and –(e)ba (-be, -bê and –bilê). For the copulative verb stem –se the tag VCOP_neg on level 2 is used, as is the case for the verb stem –be (<-ba) when it is used in the negative form.

Examples:

ke na le	VCOP_nil
ge e le marega	VCOP_nil
ge a se gona	VCOP_neg
ya ba selemo	VCOP_nil