README.md

Lexicon Annotation Guideline

This guideline outlines the tagsets and the annotation scheme that are used to manually annotate the lexicon. All lexicon annotations that are submitted to //src/analyzer/lexicon should follow this guideline.

Directory structure

Lexicon files are grouped into two:

• Base lexicon files that are provided by the Project Steward(s) in base lexicon directory (//src/analyzer/lexicon/base).
• Contrib lexicon files that are submitted by the contributors to this library and reviewed by the Project Steward(s) upon submission in contrib lexicon directory (//src/analyzer/lexicon/contrib).

Base lexicon contains 66 files with annotations on 47,202 words. Lexicon entries are grouped into different files according to their annotated coarse and fine part-of-speech tag (e.g. adjective_jj.tsv contains the annotations for base adjectival root forms; see Annotating part-of-speech for the definition of coarse and fine part-of-speech).

Structure of a lexicon file

Lexicon files are TSVs (tab separated values). They should be structured according to the following conventions:

• Lexicon files should use '.tsv' file extension.
• Lexicon files should be named as [coarse_pos_tag]_[fine_pos_tag].tsv, where [coarse_pos_tag] and [fine_pos_tag] stands for the annotated part-of-speech of the lexicon entries that can be found in that file.
• Lexicon file that is named as [coarse_pos_tag]_[fine_pos_tag].tsv should only contain annotations for lexicon entries that are annotated for [fine_pos_tag].
• Submitted lexicon files should end with an empty line (\n).
• Each lexicon file should contain a header in its first row.
• Headers should have 5 columns (tag, root, morphophonemics, features and is_compound exactly in this order).
• Rows beneath the header should include annotated lexicon entries.
• Every lexicon entry annotation should have 5 columns (each of which should contain the annotated value for the respective field in the header).

Definition of the annotated fields are as follows:

• tag (string): part-of-speech tag of the root form (see Annotating part-of-speech).
• root (string): stem of the word (see Annotating root form).
• morphophonemics (string): morphophonemic irregularities of the root form, which determines the surface form of the root after affixation of suffix morphemes (see Annotating morphophonemic irregularities).
• features (string): morphological feature tags, which are used for marking optional morphosemantic or morphosyntactic tags to be used in building a morphological analyzer for Turkish (see Annotating features).
• is_compound (bool): marks nominal root forms that end with a Turkish compounding marker +SH (see Annotating compound roots).

Annotating part-of-speech

Every lexicon entry should have the part-of-speech annotated in the tag field.

To annotate part-of-speech;

1. decide on the root form by referring to Annotating root form.
2. refer to the table of annotation tags and their descriptions and use the annotation tag that applies to the root form as its part-of-speech annotation.

Coarse part-of-speech tagset

Coarse part-of-speech tagset is a broad classification of the lexical categories. It is not used in lexicon annotations but provided here as a reference syntactic categorization to help develop a Turkish morphotactic model (since agglutination patterns for each coarse part-of-speech vary drastically).

Below listed 15 tags compose our coarse part-of-speech tagset.

Coarse Tag	Description
ADJ	Adjective
ADP	Adposition
ADV	Adverb
AFFIX	Affix
CONJ	Conjunct
DET	Determiner
EXS	Existential
NOUN	Noun
NUM	Number
ONOM	Onomatopoeia
PRON	Pronoun
PRT	Particle
PUNCT	Punctuation
VERB	Verb
X	Other

Fine part-of-speech tagset

Fine part-of-speech tagset defines the lexical categories that can occur in morphological analyses that are output by a morphological analyzer implementation for Turkish. It is not used in the lexicon annotations.

This tagset provides a finer distinction of syntactic subcategorization in comparison to coarse part-of-speech tagset. Each fine part-of-speech category can be mapped to one of the coarse part-of-speech categories. Therefore, there is a one-to-many mapping from coarse part-of-speech tags to fine part-of-speech.

Below listed 47 tags compose our fine part-of-speech tagset.

Note that some tags are specified as derived categories (marked as Yes under Is Derived Category? column). Morphologically derived categories can only occur in a morphological analysis as a result of affixation of a derivational morpheme, therefore no cross classified lexicon will contain entries with those fine part-of-speech tags. Such morphological derivations and inventory of derivational morphemes are defined as part of the morphotactic model.

Coarse Tag	Fine Tag	Is Derived Category?	Description
ADJ	JJ	No	Adjective
ADJ	VJ	Yes	Verb in participle form
ADP	IN	No	Postposition
ADV	CRB	Yes	Converb
ADV	RB	No	Adverb
ADV	WRB	No	Interrogative adverb
AFFIX	PFX	No	Prefix
CONJ	CC	No	Coordinating conjunction
DET	DT	No	Determiner
DET	PDT	No	Predeterminer
DET	WDT	No	Wh-determiner
EXS	EX	No	Existential
NOUN	ADD	No	Electronic address (e-mail or URL)
NOUN	NN	No	Common noun
NOUN	NNP	No	Proper noun
NOUN	VN	Yes	Verbal noun (head of a noun clause)
NUM	CD	No	Cardinal number
ONOM	DUP	No	Onomatopoeic
PRON	PRD	No	Demonstrative pronoun
PRON	PRF	Yes	Morphologically derived pronoun
PRON	PRI	No	Indefinite pronoun
PRON	PRP	No	Personal pronoun
PRON	PRP$	No	Possessive pronoun
PRON	PRR	No	Reflexive pronoun
PRON	WP	No	Wh-pronoun
PRT	EP	No	Final particle
PRT	OP	No	Coordinative article
PRT	RPC	No	Clitic particle
PRT	RPNEG	No	Negation particle
PRT	RPQ	No	Question particle
PUNCT	.	No	Terminal punctuation such as .!?
PUNCT	,	No	Comma and comma-like punctuation
PUNCT	:	No	Colon and semi-colon
PUNCT	(	No	Left bracket punctuation
PUNCT	)	No	Right bracket punctuation
PUNCT	``	No	Open quotation mark and similar punctuation
PUNCT	'	No	Close quotation mark and other similar punctuation
PUNCT	-	No	Hyphen, dashes, and similar punctuation
VERB	NOMP	No	Nominal predicate
VERB	VB	No	Verb
X	FW	No	Foreign word whose meaning is not known and cannot be inferred
X	GW	No	Word parts separated due to bad tokenization
X	LS	No	List symbols
X	NFP	No	Non-final punctuation, including emoticons and multi-symbol tokens
X	SYM	No	Symbol
X	UH	No	Interjection or hesitation
X	XX	No	Total garbage

Annotation tagset

Below listed 75 tags are used in lexicon annotations. Every lexicon entry should have one of the part-of-speech tags listed under the Annotation Tag column in its tag field.

In order to keep the lexicon size minimal we introduce a concept called cross classification of lexicon entries.

It assumed that for every lexicon entry that is annotated with certain annotation tags, the final lexicon should contain identical lexicon entries to that annotated entry which only differ in part-of-speech (below table presents the parts-of-speech which can be cross classified under the Cross Classifies As column). Therefore, our lexicon is compressed. To obtain the fully expanded lexicon a pre-processing stage is required in which new lexicon entries are added for cross classification parts-of-speech (e.g. a lexicon entry annotation which has the annotation tag PDT triggers creation of two lexicon entries where one has the tag NOMP and the other has the tag PDT, see //src/analyzer/lexicon/parser.py).

Coarse Tag	Annotation Tag	Cross Classifies As	Description
ADJ	JJ	JJ, NN, NOMP, PRI, RB	Adjective
ADJ	JJN	JJ, NN, NOMP	Adjective which can also be used as noun and nominal predicate
ADP	IN	IN, NN, NOMP	Postposition
ADV	RB	RB	Adverb
ADV	RB-TEMP	NN-TEMP, NOMP, RB	Temporal adverb
ADV	WRB	NOMP, WRB	Interrogative adverb
AFFIX	PFX	PFX	Prefixes
CONJ	CC	CC	Coordinating conjunction
DET	DT	DT, NOMP, PRI	Determiner
DET	PDT	NOMP, PDT	Predeterminer
DET	WDT	NOMP, PRI, WDT	Wh-determiner
EXS	EX	EX, NOMP-CASE-BARE	Existential
NOUN	ADD	ADD, NOMP-WITH-APOS	Electronic address (e-mail or URL)
NOUN	NN	NN, NOMP	Common noun
NOUN	NN-ABBR	NN, NOMP-WITH-APOS	Common noun abbreviation whose root might be separated from suffixes with an apostrophe
NOUN	NN-ABBR-APOS	NN, NOMP-APOS	Common noun abbreviation whose root might be separated from suffixes with an apostrophe only when used as a nominal predicate
NOUN	NN-TEMP	NN	Common noun that denotes temporality
NOUN	NNP	NNP, NOMP-WITH-APOS	Proper noun
NOUN	NNP-ABBR	NNP, NOMP-WITH-APOS	Proper noun abbreviation
NUM	CD	CD, NN, NOMP-WITH-APOS	Cardinal number
NUM	CD-DIST	NN, NOMP-WITH-APOS	Distributive number
NUM	CD-ORD	NN, NOMP-WITH-APOS	Ordinal number
ONOM	DUP	DUP	Onomatopoeic
PRON	PRD	NOMP, PRD	Demonstrative pronoun
PRON	PRD-PNON	NOMP-PNON, PRD	Demonstrative pronoun whose root is marked for person-number and none possessiveness
PRON	PRD-PNPOSS	NOMP-PNPOSS, PRD	Demonstrative pronoun whose root is marked for person-number and whose case markers are always realized as if it is marked for posessiveness
PRON	PRI	NOMP, PRI	Indefinite pronoun
PRON	PRP	NOMP-PN, PRP	Personal pronoun
PRON	PRP-CASE	NOMP-CASE-MARKED, PRP	Personal pronoun whose root is marked for person-number, none possessiveness and case
PRON	PRP-IRR	NOMP-PNON, PRP	Personal pronoun whose root is marked for person-number and none posessiveness, which cannot be inflected for dative case
PRON	PRP$	NOMP-PNON, PRP$	Possessive pronoun
PRON	PRR	NOMP, PRR	Reflexive pronoun
PRON	WP	NOMP, WP	Wh-pronoun
PRT	EP	EP	Final particle
PRT	OP	OP	Coordinative article
PRT	RPC	RPC	Clitic particle
PRT	RPNEG	NOMP-CASE-BARE, RPNEG	Negation particle
PRT	RPQ	NOMP-CASE-BARE, RPQ	Question particle
PUNCT	PUNCT-6	.	Terminal punctuation such as .!?
PUNCT	PUNCT-4	,	Comma and comma-like punctuation
PUNCT	PUNCT-7	:	Colon and semi-colon
PUNCT	PUNCT-2	(	Left bracket punctuation
PUNCT	PUNCT-3	)	Right bracket punctuation
PUNCT	PUNCT-8	``	Open quotation mark and similar punctuation
PUNCT	PUNCT-1	'	Close quotation mark and other similar punctuation
PUNCT	PUNCT-5	-	Hyphen, dashes, and similar punctuation
VERB	NOMP	NOMP	Nominal predicate
VERB	NOMP-APOS	NOMP	Nominal predicate whose root is always separated from suffixes with an apostrophe
VERB	NOMP-CASE-BARE	NOMP	Nominal predicate whose root is caseless but marked for person-number and none possessiveness
VERB	NOMP-CASE-MARKED	NOMP	Nominal predicate whose root is marked for person-number, none possessiveness and case
VERB	NOMP-PN	NOMP	Nominal predicate whose root is marked for person-number
VERB	NOMP-PNON	NOMP	Nominal predicate whose root is marked for person-number and none possessiveness
VERB	NOMP-PNPOSS	NOMP	Nominal predicate whose root is marked for person-number and whose case markers are always realized as if it is marked for posessiveness
VERB	NOMP-WITH-APOS	NOMP	Nominal predicate whose root might be separated from suffixes with an apostrophe
VERB	VB-HL-AR-DHR	VB	Verb that takes +Hl as passive, +Ar as aorist tense and +DHr as causative marker
VERB	VB-HL-AR-HR	VB	Verb that takes +Hl as passive, +Ar as aorist tense and +Hr as causative marker
VERB	VB-HL-AR-HT	VB	Verb that takes +Hl as passive, +Ar as aorist tense and +Ht as causative marker
VERB	VB-HL-AR-NO	VB	Verb that takes +Hl as passive and +Ar as aorist tense marker, but does not take a causative marker
VERB	VB-HL-AR-T	VB	Verb that takes +Hl as passive, +Ar as aorist tense and +t as causative marker
VERB	VB-HL-HR-DHR	VB	Verb that takes +Hl as passive, +Hr as aorist tense and +DHr as causative marker
VERB	VB-HL-HR-NO	VB	Verb that takes +Hl as passive and +Hr as aorist tense marker, but does not take a causative marker
VERB	VB-HL-HR-T	VB	Verb that takes +Hl as passive, +Hr as aorist tense and +t as causative marker
VERB	VB-HN-AR-DHR	VB	Verb that takes +Hn as passive, +Ar as aorist tense and +DHr as causative marker
VERB	VB-HN-HR-DHR	VB	Verb that takes +Hn as passive, +Hr as aorist tense and +DHr as causative marker
VERB	VB-HN-HR-NO	VB	Verb that takes +Hn as passive and +Hr as aorist tense marker, but does not take a causative
VERB	VB-HN-HR-T	VB	Verb that takes +Hn as passive, +Hr as aorist tense and +t as causative marker
VERB	VB-ON-OR-DHR	VB	Verb that takes +Hn as passive, +r as aorist tense and +DHr as causative marker
VERB	VB-ON-OR-T	VB	Verb that takes +Hn as passive, +r as aorist tense and +Dt as causative marker
X	FW	FW	Foreign word whose meaning is not known and cannot be inferred
X	GW	GW	Word parts separated due to bad tokenization
X	LS	LS	List symbols
X	NFP	NFP	Non-final punctuation, including emotico nsand multi-symbol tokens
X	SYM	SYM	Symbol
X	UH	UH	Interjection or hesitation
X	XX	XX	Total garbage

Annotating root form

Every lexicon entry should have the root form annotated in the root field.

To annotate root forms;

1. decide on the part-of-speech of the root by referring to Annotating part-of-speech.
2. strip off all the suffixes (inflectional and derivational morphemes) from the word form, given a morphotactic model that defines the inflections and derivations for part-of-speech and the corresponding morpheme inventory.
3. remaining string is the root form.
4. if the root form ends with a compounding marker +SH leave it as a part of the root form annotation (e.g. the word ahçıbaşınınki should have ahçıbaşı as its root form annotation, not ahçıbaş).

Annotating morphophonemic irregularities

Before you read this section it is strongly advised to refer to Oflazer, K. (1994). Two-level description of Turkish morphology. Literary and linguistic computing, 9(2), 137-148 for an outline of the Turkish morphophonemic processes.

Every lexicon entry should have the morphophonemic irregularities annotated in the morphophonemics field.

To annotate morphophonemic irregularities;

1. go through the list of rules presented in below sections and decide on the morphophonemic irregularity annotation string by referring to those that apply to the root form.
2. if more than one of these rules apply to the root form, what all applying annotation rules dictate should be incorporated in the final morphophonemic annotation string.
3. if none of these rules apply to the root form, use ~ as the morphophonemic irregularity annotation string.

Voicing exceptions

The voicing exception applies to roots whose final voiceless consonant fails to get voiced in spite of the affixation of a suffix that starts with a vowel. This exception applies only to sounds that are [-voiced][+plosive] (e.g. { p, t, k, t, ʃ }). Voiceless continuants { f, s, ʃ } never undergo voicing anyway.

Roots ending with K annotation mark up

Root final voiceless plosive k is always assumed to realize as ğ when a suffix starting with a vowel is affixed. Sometimes root final voiceless plosive k realizes as g instead of ğ only if it is preceded by n.

If the root ends with voiceless plosive k which is left unchanged (and not realized as ğ or g) after a suffix starting with a vowel is affixed, then the root final velar stop k is annotated with the markup K.

For the sake of consistency, even there are no suffixes which start with a vowel that can be affixed to the root in the affix inventory of the morphotactic model for the annotated part-of-speech of the root, this rule still applies (e.g. as seen in the adjectival examples below).

Coarse Tag	Root	Morphophonemic Annotation	Example
ADJ	ak	aK	ak
ADJ	tok	toK	tok
NOUN	mercanköşk	mercanköşK	mercanköşk-e
NOUN	meşk	meşK	meşk-i
NOUN	şark	şarK	şark-ı
NOUN	türk	türK	türk-e
VERB	bük	büK	bük-er
VERB	kalk	kalK	kalk-ar
VERB	sok	soK	sok-ar

Roots ending with ~ annotation mark up

Root final voiceless plosive p (or t) is always assumed to be left unchanged after a suffix starting with a vowel is affixed.

If the root ends with the voiceless plosive p (or t) which is realized as voiced b (or resp. d) when a suffix starting with a vowel is affixed, then the root final voiceless plosive p (or t) should be annotated with a succeeding ~ annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	dolap	dolap~	dolab-ın
NOUN	mikrop	mikrop~	mikrob-u
NOUN	kanat	kanat~	kanad-ının
NOUN	tehdit	tehdit~	tehdid-i
VERB	bahset	bahset~	bahsed-iyor
VERB	git	git~	gid-ecek

Roots ending with Ç annotation mark up

Root final voiceless plosive ç is always assumed to realize as voiced c when a suffix starting with a vowel is affixed.

If the root ends with the voiceless plosive ç which is left unchanged (and not realized as c) after a suffix starting with a vowel is affixed, then the root final voiceless plosive ç is annotated with the markup Ç.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	göç	göÇ	göç-e
NOUN	sandviç	sandviÇ	sandviç-in
NOUN	suç	suÇ	suç-u
NUM	üç	üÇ	üç-er
VERB	aç	aÇ	aç-ıl
VERB	seç	seÇ	seç-er

Vowel harmony exceptions

Palatalization of final l

The lateral l has allophones when it occurs in the root final position after back vowels.

Velar l (back)	Palatal l (front)
sol	sol (anahtarı)
okul	usul
araba	kalp
bor	golf

When roots that have a palatalized root final l are followed by a suffix that starts with a vowel, the surface form of the suffixes are resolved as if the last syllable of the root has a front vowel.

Roots ending with % annotation in the last vowel position

Certain roots that have the back rounded vowel o as its last vowel followed by a l cause the first back vowel of the affixed suffix to be realized as a front vowel.

The last vowel of those roots should be annotated with the markup %.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	mentol	ment%l	mentol-lü
NOUN	metropol	metrop%l	metropol-de
NOUN	monokl	monok%l	monokol-e

Roots ending with { annotation in the last vowel position

Certain roots that have the back unrounded vowel a as its last vowel followed by a l cause the first back vowel of the affixed suffix to be realized as a front vowel.

The last vowel of those roots should be annotated with the markup {.

Coarse Tag	Root	Morphophonemic Annotation	Example
ADJ	biyomedikal	biyomedik{l	biyomedikal-de
NOUN	ideal	ide{l	ideal-i
NOUN	ihtimal	ihtim{l	ihtimal-iyle

In addition, if the root ends with a cluster that contains a back vowel and { r, t, b, d }, same phenomena could be observed. The last vowel of such roots should also be annotated with the markup {.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	dikkat	dikk{t	dikkat-i
NOUN	harf	hr{f	harf-e
NOUN	hat	h{t~"	had-dinin
NOUN	saat	sa{t	saat-e

Roots ending with } annotation in the last vowel position

Certain roots that have the back rounded vowel u as its last vowel followed by a l cause the first back vowel of the affixed suffix to be realized as a front vowel.

The last vowel of those roots should be annotated with the markup }.

Coarse Tag	Root	Morphophonemic Annotation	Example
ADJ	mesul	mes}l	mesul-ü
NOUN	ampul	amp}l	ampul-e
NOUN	duhul	duh}l	duhul-de

Roots ending with [ annotation in the last vowel position

Certain roots that have the long â as its last vowel followed by a l cause the first back vowel of the affixed suffix to be realized as a front vowel.

The last vowel of those roots should be annotated with the markup [.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	eşkâl	eşk[l	eşkâl-i
NOUN	hâl	h[l	hâl-den
NOUN	işkâl	işk[l	işkâl-inin

Vowel closing

For certain verbal roots, root final front unrounded open vowel e is realized as close i when a suffix starting with a vowel is affixed.

The final front unrounded vowel e of those roots should be annotated with the E annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
VERB	de	dE	d-i-yecek
VERB	ye	yE	y-i-yor

Epenthesis

Last vowel of certain root forms are dropped when a suffix that starts with a vowel or a consonant that drops gets affixed.

The final vowel of those roots should be annotated with a succeeding ? annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	böğür	böğü?r	böğr-üme
NOUN	burun	buru?n	burn-um
NOUN	isim	isi?m	ism-e
VERB	çağır	çağı?r	çağr-ıldı

Gemination

For some words, root final consonant { b, c, d, f, k, l, m, n, r, t, z } is duplicated when a suffix starting with a vowel is affixed.

The final consonant of those roots should be annotated with a succeeding " annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
ADJ	muhil	muhil"	muhil-l-i
ADJ	muhip	muhip~"	muhib-b-in
NOUN	af	af"	af-f-ın
NOUN	hak	hak"	hak-k-a

Roots ending with -su

Certain nominal root forms that end with su alter the initial consonant of some suffixes (e.g. 3rd person possessive and genitive case inflection morphemes) when they are affixed.

Root final su of those root forms should be annotated with a succeeding ^ annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	akarsu	akarsu^	akarsu-yun
NOUN	yuzsu	yuzsu^	yuzsu-yu
NOUN	su	su^	su-yu

Foreign words and code switching

In case of code switching, foreign words could be used in Turkish sentences, while preserving their surface form and getting inflected according to the part-of-speech that they hold within the context.

Foreign word stems should be annotated by enclosing the last syllable of their Turkish pronunciation enclosed in * annotation markup.

Coarse Tag	Root	Morphophonemic Annotation	Example
NOUN	ABD	ABD*e*	ABD-de
NOUN	single	single*ıl*	single-ının

Annotating features

Certain lexicon entry annotations must be annotated with morphosyntactic and morphosemantic feature tags to represent the features that are intrinsically carried by the root itself without affixation of any inflectional or derivational morpheme.

Every lexicon entry should have features annotated in the features field. Every feature should be annotated in the form of +[feature_category=feature_tag].

To annotate features;

1. check whether annotated part-of-speech of the lexicon entry matches one of the part-of-speech listed in the tables for required and optional features (note that set of part-of-speech tags that require features annotation in both lists are disjoint).
2. if so, annotate the lexicon entry with corresponding feature tags that apply for the root.
3. if annotated part-of-speech of the lexicon entry does not match any of the part-of-speech listed in the tables for required and optional features, use ~ as the features annotation string.

Note that if a lexicon entry is required to be annotated with more than one feature, all applicable features should be annotated in sequence, in the form of +[feature_category_1=feature_tag_1]+[feature_category_2=feature_tag_2] (e.g. lexicon entries that are annotated with PRP-CASE part-of-speech tag should take three feature tags in sequence, where the first feature is chosen from the set of features tags { A1sg, A2sg, A3sg, A1pl, A2pl, A3pl } and the second chosen from { Pnon }, and so on).

Below listed 31 feature tags compose our features tagset.

Feature Category	Feature Tag	Description
PersonNumber	A1sg	1st person singular
PersonNumber	A2sg	2nd person singular
PersonNumber	A3sg	3rd person singular
PersonNumber	A1pl	1st person plural
PersonNumber	A2pl	2nd person plural
PersonNumber	A3pl	3rd person plural
Case	Abl	Ablative case marked
Case	Acc	Accusative case marked
Case	Dat	Dative case marked
Case	Gen	Genitive case marked
Case	Ins	Instrumental case marked
Case	Loc	Locative case marked
ComplementType	CAbl	(Postposition has) ablative case marked complement
ComplementType	CAcc	(Postposition has) accusative case marked complement
ComplementType	CBare	(Postposition has) caseless complement
ComplementType	CDat	(Postposition has) dative case marked complement
ComplementType	CFin	(Postposition has) finite complement
ComplementType	CGen	(Postposition has) genitive case marked complement
ComplementType	CIns	(Postposition has) instrumental case marked complement
ComplementType	CNum	(Postposition has) numeric complement
ConjunctionType	Adv	Adverbial conjunction
ConjunctionType	Coor	Coordinating conjunction
ConjunctionType	Par	Parallel conjunction
ConjunctionType	Sub	Subordinating conjunction
DeterminerType	Def	Definitive (determiner)
DeterminerType	Dem	Demonstrative (determiner)
DeterminerType	Dir	Directional (determiner)
DeterminerType	Ind	Indefinite (determiner)
Emphasis	True	Emphasis
Possessive	Pnon	None possessive
Temporal	True	Temporal

Required features

All lexicon entries that are annotated with one of the below listed part-of-speech tags are required to be annotated with the corresponding feature tag(s).

Coarse Tag	Annotation Tag	Required Features
ADP	IN	+[ComplementType=CAbl|CAcc|CBare|CDat|CFin|CGen|CIns|CNum]
ADV	RB-TEMP	+[Temporal=True]
CONJ	CC	+[ConjunctionType=Adv|Coor|Par|Sub]
DET	DT	+[DeterminerType=Def|Dem|Dir|Ind]
NOUN	NN-TEMP	+[Temporal=True]
PRON	PRD-PNON	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]
PRON	PRD-PNPOSS	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]
PRON	PRP	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]
PRON	PRP-CASE	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]+[Case=Acc|Abl|Dat|Gen|Ins|Loc]
PRON	PRP-IRR	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]
PRON	PRP-PRP$	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]
VERB	NOMP-CASE-BARE	+[PersonNumber=A3sg]+[Possessive=Pnon]+[Bare]
VERB	NOMP-CASE-MARKED	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]+[Case=Acc|Abl|Dat|Gen|Ins|Loc]
VERB	NOMP-PN	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]
VERB	NOMP-PNON	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]+[Possessive=Pnon]
VERB	NOMP-PNPOSS	+[PersonNumber=A1sg|A2sg|A3sg|A1pl|A2pl|A3pl]

Optional features

All lexicon entries that are annotated with one of the below listed part-of-speech tags might be annotated with the corresponding feature tag(s), if and only if the feature tag applies to the root.

Coarse Tag	Annotation Tag	Optional Features
ADJ	JJ	+[Emphasis=True]
ADJ	JJN	+[Emphasis=True]
ADV	RB	+[Emphasis=True|Temporal=True]

Annotating compound roots

Certain Turkish root forms end with a compounding marker +SH, which is ambiguous with a 3rd person possessive marker (e.g. common noun: atbalığı, proper noun: Kırklareli).

Defining the morphotactics for those cases are particularly challenging. For example, if the word is a common noun which is inflected for person-number, the person-number morpheme occurs before the compounding marker which is essentially part of the root form. While implementing a morphological analyzer, this phenomena violates the assumption that all suffixes are sequentially affixed to the root.

Every lexicon entry should have this phenomena annotated in the is_compound field.

To annotate roots that end with a compounding marker;

1. is_compound field should be annotated as TRUE if the root form ends with a compounding marker (otherwise, FALSE).
2. the morphophonemic irregularity annotation string in the morphophonemics field should be the root form without the final compounding marker (e.g. the word atbalığı should have atbalık as its morphophonemics annotation).
3. if any of the morphophonemics processes that are described in the Annotating morphophonemic irregularities applies to morphophonemic annotation string yielded by (2), they should also be represented in the morphophonemics annotations (e.g. adamkökü should have adamköK as its morphophonemics annotation, not adamkök).
4. lexicon entries for root forms that do not end with a compounding marker should follow the rules defined in Annotating morphophonemic irregularities.