Morphological Feature Annotation for Tamil

Annotators should add all relevant features that are morphologically visible and useful for UD annotation, although the examples in the documentation provide only the feature being discussed in each respective example.

1. PronType

PronType is used for pronouns, determiners, pronominal adverbs, and related forms.

Feature values used in Tamil

PronType=Prs   personal pronoun
PronType=Dem   demonstrative
PronType=Int   interrogative
PronType=Ind   indefinite
PronType=Neg   negative indefinite
PronType=Rcp   reciprocal
PronType=Rel   relative
PronType=Tot   total / universal

Example:

அவன் யார் ?
avaṉ yār ?
he   who
Who is he?

யார்    PRON    PronType=Int

2. Deixis

Deixis marks the relative location encoded in demonstratives. Tamil has a three-way deictic distinction:

இ-  proximate / near speaker
உ-  medial / near hearer or intermediate
அ-  distal / remote

Feature values used in Tamil

Deixis=Prox   proximate
Deixis=Med    medial
Deixis=Remt   remote / distal

Example:

உவன் வந்தான் .
uvaṉ vantāṉ
that.person.MED came
“That person came.”

Annotation:

உவன்    PRON    PronType=Dem|Deixis=Med|Gender=Masc|Number=Sing|Person=3

Possible Tamil mapping:

3. NumType

NumType marks the type of numeral.

Feature values used in Tamil

NumType=Card   cardinal number
NumType=Ord    ordinal number
NumType=Dist   distributive
NumType=Frac   fraction
NumType=Mult   multiplicative
NumType=Range  numerical range
NumType=Sets   set/group numeral

Example:

இரண்டு  நாய்கள் வந்தன .
iraṇṭu nāykaḷ vantaṉa
two    dogs   came

Annotation:

இரண்டு    NUM    NumType=Card|NumForm=Word

4. NumForm

NumForm marks how the number is written.

Feature values used in Tamil

NumForm=Word    written as a word
NumForm=Digit   written as digits
NumForm=Roman   written as Roman numerals

Example:

2 நாய்கள்  வந்தன
2 nāykaḷ vantaṉa
Two dogs came

Annotation:

2    NUM    NumType=Card|NumForm=Digit

5. Poss

Poss=Yes is used for possessive forms.

Example:

அவனுடைய புத்தகம் உள்ளது .
avaṉuṭaiya puttakam uḷḷatu
his book exists
“His book is there.”

Annotation:

அவனுடைய    PRON    PronType=Prs|Poss=Yes|Gender=Masc|Number=Sing|Person=3

6. Reflex

Reflex=Yes is used when a pronoun refers back to the subject of the clause.

Example:

அவன் தன்னை        பார்த்தான்
avaṉ taṉṉai       pārttāṉ
he   himself-ACC  saw
He saw himself

Annotation:

தன்னை    PRON    PronType=Prs|Reflex=Yes|Case=Acc|Number=Sing|Person=3

Do not mark every occurrence of தான் as reflexive. It may also function as an emphatic clitic.

7. Gender

Gender marks gender where it is morphologically or semantically relevant.

Feature values used in Tamil

Gender=Masc   masculine
Gender=Fem    feminine
Gender=Neut   neuter / non-rational
Gender=Com    common / honorific-human

Example:

அவள் வந்தாள் .
avaḷ vantāḷ
she came
“She came.”

Annotation:

அவள்       PRON       PronType=Prs|Gender=Fem|Number=Sing|Person=3
வந்தாள்      VERB      Tense=Past|VerbForm=Fin|Gender=Fem|Number=Sing|Person=3

8. Animacy

Animacy can be used to capture Tamil rational/non-rational distinctions.

Recommended Tamil values

Animacy=Anim   animate
Animacy=Inan   inanimate

Example:

பையன் வந்தான் .
paiyaṉ vantāṉ
boy came
“The boy came.”

Annotation:

பையன்    NOUN    Animacy=Anim

9. Number

Number marks singular or plural.

Main Tamil values

Number=Sing
Number=Plur

Example:

பையன்கள்   வந்தார்கள் .
paiyaṉkaḷ vantārkaḷ
boys came
“The boys came.”

Annotation:

பையன்கள்    NOUN    Number=Plur

10. Case

Case marks case on nouns, pronouns, proper nouns, and some nominalised forms.

Common Tamil values

The following table lists common Tamil case values, possible suffixes, and examples. The exact surface form may vary because of sandhi, phonological changes, and stem alternations.

Example:

குமார் ராமனை பார்த்தான் .
kumār rāmaṉai pārttāṉ
Kumar Raman-ACC saw
“Kumar saw Raman.”

Annotation:

ராமனை    PROPN    Case=Acc|Animacy=Hum|Gender=Masc|Number=Sing

Note: Tamil case suffixes are normally not tokenised separately. They are represented as morphological features on the nominal token.

Nowadays, case markers are also sometimes written separately as tokens. In such cases, they should be carefully analysed and annotated.

Example:

சாவி மூலம் கதவைத் திறந்தான்
cāvi moolam katavait tiṟantāṉ
key using   door-ACC open-PST-3SG.M
He opened the door with a key

Possible annotation:

ூலம்     ADP       Case=Ins

11. Definite

Definite may be used when indefiniteness or specificity is overtly marked.

Possible values

Definite=Ind    indefinite
Definite=Spec   specific
Definite=Def    definite, if clearly marked

Example:

ஒரு குச்சியைக்   கொண்டு வா
oru kucciyai  koṇṭu  vā
one stick-ACC bring  come
Bring a stick

Annotation:

ஒரு    NUM/DET    NumType=Card|Definite=Ind

Note: Tamil does not have articles like English. Use Definite cautiously, mainly where forms such as ஒரு, ஏதேனும், or குறிப்பிட்ட clearly mark indefiniteness or specificity.

12. Degree

Degree is used for adjectives or adverbs when comparison or superlative meaning is encoded.

Possible values

Degree=Pos
Degree=Cmp
Degree=Sup

Example:

ரோஜா மல்லிகையைவிட     அழகு
rōjā mallikaiyaiviṭa  aḻaku
rose jasmine-ACC than beautiful
Rose is more beautiful than jasmine

Annotation:

அழகு    ADJ    Degree=Cmp

Note: Tamil comparison is often expressed syntactically using விட “than”. Therefore, Degree=Cmp should be used only if the treebank policy treats the adjective as comparative in that construction.

13. AdpType

AdpType marks the type of adposition.

Tamil value

AdpType=Post

Example:

அவன்போல    பேசினான்
அவன்  போல  பேசினான்
avaṉ  pōla  pēciṉāṉ
he    like  spoke
He spoke like him

Annotation:

போல    ADP    AdpType=Post

Note: Many Tamil ADP-like forms may also function as nouns, adverbs, verbs, or subordinators. Assign ADP only after checking the context.

14. VerbForm

VerbForm marks the form of a verb.

Values used in Tamil

VerbForm=Fin    finite verb - முற்றுவினை
VerbForm=Inf    infinitive - செய்ய எச்சம்
VerbForm=Conv   converb - verbal participle / செய்து எச்சம்
VerbForm=Vnoun  verbal noun - தொழிற்பெயர்

Example:

குமார்  வேலை முடிக்கவேண்டும்
குமார்  வேலை முடிக்க     வேண்டும்
kumār vēlai muṭikka   vēṇṭum
Kumar work  finish-INF must
“Kumar must finish the work.”

Annotation:

முடிக்க    VERB    VerbForm=Inf

Note: Traditional Tamil வினையெச்சம் forms such as செய்து, வந்து, பார்த்து may be annotated in UD as VerbForm=Conv. In glosses for Tamil annotators, these can be labelled as VRP.

15. Mood

Mood marks verbal mood or modality.

Common values for Tamil

Example:

இந்த புத்தகத்தைக்     கொடுங்கள் .
inta puttakattaik koṭuṅkaḷ
this book-ACC     give-IMP.2PL
Give this book

Annotation:

கொடுங்கள்    VERB    Mood=Imp|VerbForm=Fin|Person=2|Polite=Form

Note: Tamil modality is often expressed through auxiliaries or modal predicates such as வேண்டும், முடியும், மாட்டேன், and -லாம். Do not assign mood mechanically without checking the construction.

16. Tense

Tense marks tense on finite verbs and auxiliaries.

Tamil values

Tense=Past
Tense=Pres
Tense=Fut

Example:

குமார்  இட்லியைச் சாப்பிட்டான்
kumār iṭliyaic cāppiṭṭāṉ
Kumar idli-ACC ate
Kumar ate idli

Annotation:

சாப்பிட்டான்    VERB    Tense=Past|VerbForm=Fin|Person=3|Number=Sing|Gender=Masc

17. Aspect

Aspect may be used when aspectual meaning is morphologically or constructionally visible.

Possible Tamil-relevant values

Aspect=Prog   progressive
Aspect=Perf   perfect / completive
Aspect=Hab    habitual, if clearly marked

Example:

அவன் படித்துக்கொண்டிருந்தான்
அவன் படித்து     கொண்டு   இருந்தான்
avaṉ paṭittuk koṇṭ      iruntāṉ
he   study     keep.VRB be-PST.3SG.M
He was studying

Possible annotation:

கொண்டு      AUX    Aspect=Prog|VerbForm=Conv
இருந்தான்    AUX     Tense=Past|VerbForm=Fin|Person=3|Number=Sing|Gender=Masc

18. Voice

Voice marks voice distinctions.

Tamil values

Voice=Act
Voice=Pass
Voice=Cau

Example:

குமார்  அப்பாவால்    அடிக்கப்பட்டான்
குமார்  அப்பாவால்    அடிக்கப்    பட்டான்
kumār appāvāl    aṭikkap   paṭṭāṉ
Kumar father-INS beat-INF  undergo-PST-3SG.M
Kumar was beaten by his father

Annotation:

பட்டான்    AUX    Voice=Pass|Tense=Past|VerbForm=Fin|Person=3|Number=Sing|Gender=Masc

19. Polarity

Polarity marks positive or negative polarity.

Values

Polarity=Neg
Polarity=Pos

Example:

நான் விடமாட்டேன்
நான் விட     மாட்டேன்
nāṉ viṭa    māṭṭēṉ
I leave-INF will.not
I will not leave

Annotation:

மாட்டேன்    AUX    Polarity=Neg|Person=1|Number=Sing|VerbForm=Fin

20. Person

Person marks first, second, or third person.

Values

Person=1
Person=2
Person=3

Example:

நான்   வருகிறேன்
nāṉ   varukiṟēṉ
I     come-PRS-1SG
I am coming

Annotation:

நான்         PRON    PronType=Prs|Person=1|Number=Sing
வருகிறேன்     VERB    Tense=Pres|VerbForm=Fin|Person=1|Number=Sing

21. Polite

Polite marks politeness, respect, or honorific status.

Possible Tamil values

Polite=Form   formal / honorific
Polite=Infm   informal
Polite=Elev   elevated honorific title

Example:

மேதகு      தமிழக      முதலமைச்சர்       கொடி ஏற்றினார்.
mētaku     tamiḻaka  mutalamaiccar   koṭi ēṟṟiṉār
honourable Tamil.Nadu chief.minister flag hoist-PST.3SG.COM
‘The Honourable Chief Minister of Tamil Nadu hoisted the flag.

Annotation:

மேதகு          ADJ    Polite=Elev

Note: Tamil honorific marking can override ordinary number/gender agreement. The annotation policy should define how to handle honorific plural forms used for a single person. Because, in the example above, it can also be written as ஏற்றினார்கள், where -கள் is a plural marker, but it also gives an additional honorific meaning.

22. Clusivity

Clusivity distinguishes inclusive and exclusive first-person plural pronouns.

Tamil values

Clusivity=In   inclusive we
Clusivity=Ex   exclusive we

Example:

நாம்     செல்வோம் .
nām     celvōm
we.INCL go-FUT-1PL
“We will go.”

Annotation:

நாம்    PRON    PronType=Prs|Person=1|Number=Plur|Clusivity=In

Suggested Tamil policy:

Note: This distinction should be verified in the corpus because actual usage may vary by dialect and register.

23. Foreign

Foreign=Yes is used for foreign words that are not integrated into Tamil.

Example:

குமார்   பாஸ்ட்டாக ஓடினான் .
kumār  pāsṭṭāka ōṭiṉāṉ
Kumar  fast-ADV ran
“Kumar ran fast.”

Annotation:

பாஸ்ட்டாக    ADV    Foreign=Yes

Note: Some loanwords such as கார் “Car”, பஸ் “Bus”, டிவி “TV”, சினிமா “Cinema” and மொபைல் “Mobile” are now well integrated in Tamil usage. Thereore, it is upto the project to decide whether to mark them foreign or not.

24. Abbr

Abbr=Yes is used for abbreviations.

Example:

தி.மு.க.   ஆட்சியில்   உள்ளது
ti.mu.ka. āṭciyil   uḷḷatu
DMK       power-LOC exist
“DMK exists in power.”

Annotation:

தி.மு.க.    PROPN    Abbr=Yes

25. Typo

Typo=Yes is used when a token contains an error that is preserved in the corpus.

Example:

அவ்ன் வந்தான் .
avṉ vantāṉ
“He came.”

Annotation:

அவ்ன்    PRON    Typo=Yes

26. PunctType

PunctType marks the type of punctuation.

Common values

PunctType=Peri   full stop
PunctType=Qest   question mark
PunctType=Excl   exclamation mark
PunctType=Comm   comma
PunctType=Quot   quotation mark
PunctType=Brck   bracket

Example:

அவன் வந்தானா ?
avaṉ vantāṉā ?
Did he come ?

Annotation:

?    PUNCT    PunctType=Qest

27. Sandhi

Sandhi is proposed as a Tamil-specific feature to record surface changes.

Proposed values

Sandhi=k  க்
Sandhi=c  ச்
Sandhi=t  த்
Sandhi=p  ப்

Tamil morphological analysis and syntactic annotation are often affected by sandhi, as sandhi may introduce restrictions on how words are segmented and interpreted. In some cases, sandhi markers may also indicate genitive-like relations, as in கதிரைக் கால் compared with கதிரை கால்.

கதிரைக்           கால்
katirai-k        kāl
chair-GEN.SANDHI leg
‘the leg of the chair’

கதிரை   கால்
katirai kāl
chair   leg

Marking sandhi is also useful because, in contemporary writing, some writers do not consistently represent sandhi. Therefore, sandhi marking can provide valuable features for studying variation in Tamil texts, including changes across time, generations, genres and writing styles.

Note: Since Sandhi is not a core universal UD feature, the treebank should define it clearly as a Tamil-specific extension.

Morphological annotation should be rich enough to support syntactic analysis. Annotators should not stop after identifying only UPOS or one obvious feature. Add all relevant features that are visible in the form and justified by the context.

© 2014–2024 Universal Dependencies contributors. Site powered by Annodoc and brat