AN INVESTIGATION OF KAZAN TATAR MORPHOLOGY

_______________

A Thesis

Presented to the

Faculty of

San Diego State University

_______________

In Partial Fulfillment

of the Requirements for the Degree

Master of Arts

in

Linguistics

_______________

by

Albina Raisovna Davliyeva

Spring 2011
 iii

Copyright © 2011

by

Albina Raisovna Davliyeva

All Rights Reserved

 iv

DEDICATION

To my grandmother Zakira who engrained in me the little Tatar that I know.

To my parents, Marsilia and Rais, who engrained in me the value of learning since
the early age.
To my niece Diana and nephew Daniel, may you fall in love with the Tatar language.
 v

ABSTRACT OF THE THESIS

An Investigation of Kazan Tatar Morphology

by
Albina Raisovna Davliyeva
Master of Arts in Linguistics
San Diego State University, 2011

This paper describes the morphological processor for parsing and generation of the
Kazan Tatar language designed and implemented using XEROX finite-state tools. Kazan
Tatar is one of the two official languages of the Republic of Tatarstan in the Russian
Federation. It is characterized by rich and complex agglutinating morphology with
phonologically determined allomorphy. This richness and complexity is defined by the
number of suffixes, morpheme functions, and recursive word formation devices. Kazan Tatar
grammar, including morphotactics and phonological processes, relevant to the project is
discussed along with the implementation details.
 vi

TABLE OF CONTENTS

PAGE

ABSTRACT ...............................................................................................................................v
LIST OF TABLES ................................................................................................................. viii
LIST OF FIGURES ................................................................................................................. ix
ABBREVIATIONS ...................................................................................................................x
ACKNOWLEDGEMENTS ..................................................................................................... xi
CHAPTER
1 INTRODUCTION .........................................................................................................1 
Objective ..................................................................................................................1 
Project Significance .................................................................................................2 
Kazan Tatar Language .............................................................................................5 
Tatar Phonetic Model .........................................................................................8 
Tatar Morphology ............................................................................................11 
Morphotactic Rules ....................................................................................15 
Recursive Concatenation ...........................................................................16 
Noun ...........................................................................................................17 
Verb............................................................................................................28 
Pronouns ....................................................................................................34 
Adjective ....................................................................................................35 
Methodology ..........................................................................................................36 
Implementation ................................................................................................37 
Lexicon ............................................................................................................39 
Morphophonological Rules ..............................................................................41 
Conclusion and Future Research ...........................................................................48 
REFERENCES ........................................................................................................................50
APPENDIX
A TATAR VERB PARADIGM ......................................................................................52
 vii

B EXAMPLES OF MORPHOLOGICAL ANALYSIS AND GENERATION .............54 

C KATMORPH XFST IMPLEMENTATION................................................................58 
 viii

LIST OF TABLES

PAGE

Table 1. Morphological Analysis and Production .....................................................................2 

Table 2. Alphabet Transliteration ..............................................................................................6 
Table 3. Front Vowels................................................................................................................8 
Table 4. Back Vowels ................................................................................................................8 
Table 5. Tatar Consonants .........................................................................................................9 
Table 6. Examples of Vowel Harmony Rules .........................................................................10 
Table 7. Morphological Feature Tags ......................................................................................13 
Table 8. Noun Derivational Suffixes .......................................................................................18 
Table 9. Tatar Noun Singular and Plural Forms ......................................................................18 
Table 10. Tatar Personal Suffixes ............................................................................................19 
Table 11. Noun Declension by Possesion ................................................................................20 
Table 12. Case Paradigm for Nouns ........................................................................................20 
Table 13a. Declension of apa ..................................................................................................27 
Table 13b. Declension of apa, 1st Person Possessor ................................................................27 
Table 13c. Declension of apa, 2nd Person Possessor ...............................................................27 
Table 13d. Declension of apa, 3rd Person Possessor ...............................................................28 
Table 14. Verb Negation ..........................................................................................................29 
Table 15. Verb Conjugation by Tense for the 3rd Person Singular ..........................................29 
Table 16. Imperative Mood Conjugation .................................................................................30 
Table 17. Verb Personal Suffixes ............................................................................................31 
Table 18. Present Tense Paradigm ...........................................................................................32 
Table 19. Verb Conjugation for the Past Tense .......................................................................33 
Table 20. Verb Conjugation for the Future Tense ...................................................................33 
Table 21. Personal Pronouns Paradigm ...................................................................................34 
Table 22. Adjective Derivational Morphemes .........................................................................35 
Table 23. Regular Expressions Notation .................................................................................44 
Table 24. Verb Paradigm of Tabarga ‘to Find’........................................................................53 
 ix

LIST OF FIGURES

PAGE

Figure 1. Morphotactic rules. ...................................................................................................15 

Figure 2. Tatar word formation. ...............................................................................................16 
Figure 3. FST architecture. ......................................................................................................38 
Figure 4. Adjective lexicon. .....................................................................................................42 
 x

ABBREVIATIONS

Aff Affirmative
Comp Comparative
DB Derivational Boundary
FSA Finite-State Automation
FSN Finite-State Network
FST Finite-State Transducer
IF Intermediate Form
IPA International Phonetic Alphabet
KaTMorph Kazan Tatar Morphological Processor
N Noun
Neg Negative
NLP Natural Language Processing
NP Noun Phrase
Obj Object
POS Part of Speech
SF Surface Form
VP Verb Phrase
UF Underlying Form
XFST Xerox Finite-State Transducer
 xi

ACKNOWLEDGEMENTS

This project is carried out thanks to many people: my professors, my family, and my
friends.
I would like to extend my deepest gratitude to my linguistics professors who have
inspired me with their passion and knowledge of linguistics as well as their dedication to
teaching (the names are listed alphabetically):
Jon Dayley
Jean Mark Gawron
Robert Malouf
Mary Ellen Rider
Gail Shuck
Robert Underhill
I owe my gratitude to my thesis committee members Robert Underhill, Robert
Malouf, and John Carroll whose encouragement, guidance and support from the initial to the
final level enabled me to develop an understanding of the subject.
I would also like to acknowledge my anthropology professors Robert McCarl and
John Ziker for raising my interest in the cultural and social aspects of linguistics.
I would like to thank Hadi Karimov, Alfina Galiahmetova, and my parents who
provided native speaker judgments. I am also thankful to Steve Milaskey, Karen Cook, Dale
Spindler, Dave Walsh, and Jared Milaskey for proofreading the manuscript.
I would like to gratefully acknowledge my supervisor Mark Morsch for being
accommodating with my school schedule.
I am thankful to Elvira, Steve, and Aliya for their moral support throughout my
academic journey. I wish to thank my fellow graduate students Gina, Brandon, Dave, Rafi,
and Paulo for providing a stimulating and fun environment to learn and grow.
Lastly, I thank all of those who supported me in any respect during the completion of
this project.
 1

CHAPTER 1

INTRODUCTION

OBJECTIVE
The goal of this project is to design and implement a Kazan Tatar Morphological
Processor (KaTMorph) that can analyze a grammatical Kazan Tatar word by annotating its
derivational and grammatical features, and can generate a well-formed Kazan Tatar word
from a specified root and grammatical features. This processor is a finite-state “transducer
that incorporates lexicon, morphotactic, and morphophonological alternations” (Beesley &
Karttunen, 2003, p. xvi); it maps properly spelled words to morphosyntactic interpretation
and vice versa. It is developed mostly for inflectional morphology of nouns and verbs;
minimal derivational morphology is included in the source code to demonstrate word
derivation in Kazan Tatar and to provide the framework for future development. KaTMorph
should accept and produce the grammatical strings for the Kazan Tatar language, and reject
and not generate ungrammatical strings. The task of the processor is to be able to take the
input forms of the type listed in column 1 of Table 1, and to produce the output forms of the
type listed in column 2.
To design a morphological processor, the following knowledge of the language
structure is required (Jurafsky & Martin, 2008):
1. lexicon, which lists word roots and morphemes
2. morphotactic rules, which explain morpheme ordering within a word
3. phonological rules that affect orthography when morphemes combine
The choice of developing the morphological processor for Kazan Tatar using finite-
state Networks (FSN) is due to the fact that concatenative morphology, characteristic of a
Turkic language, can be elegantly described by finite automata (Trost, 2003). In addition,
computing with FSN is attractive because they are computationally efficient, require
relatively little memory for storage, are known for their accuracy, and their mechanism is
language independent (Beesley & Karttunen, 2003). KaTMorph contains 28 noun, 16 verb, 6
 2

Table 1. Morphological Analysis and Production

Input Output Gloss
Parsing
miNa min-Pron+Dat to me
minem min-Pron+P1-Sg+Poss my
tapsInnar tap-Verb+Imp+P3-PL let them find
tapsIn tap-Verb+Imp+P2-Sg let him find
kolaklarnI kolak-Noun+PL+Def-Acc ears (Object)
kolaklar kolak-Noun+PL+Nom ears
matur matur-Adj beautiful
maturlIk matur-Adj+^DB-Noun+Sg+Nom beauty

Generation
min-Pron+Dat miNa to me
min-Pron+P1-Sg+Poss minem my
tap-Verb+Imp+P3-PL tapsInnar let them find
tap-Verb+Imp+P2-Sg tapsIn let him find
kolak-Noun+PL+Def-Acc kolaklarnI ears (Object)
kolak-Noun+PL+Nom kolaklar ears
matur-Adj matur beautiful
matur-Adj+^DB-Noun+Sg+Nom maturlIk beauty

pronoun, and 3 adjective roots and generates an unlimited number of inflected word forms
while occupying only 290 Kbytes1 of memory and compiles in seconds.

PROJECT SIGNIFICANCE
The project significance is determined by the general importance of morphological
knowledge for language processing by humans and computers, and by language specific
considerations. In computational linguistics, the applications of morphological analyzers are
vast. They can be used as stand-alone applications or as a preprocessing stage for other
projects or systems in natural language processing (NLP) such as part of speech tagging,
parsing, and translation.
A morphological analyzer can be used to investigate language morphology and
syntax for linguistic research and for didactic purposes. Understanding word structure is
important for comprehension of agglutinative languages because morphemes carry

1
Interestingly, this project discussion paper occupies 796 Kbytes of memory.
 3

information about grammatical relations of words in a sentence. For instance, understanding

morphological word structure of languages like Kazan Tatar is necessary in order to
disambiguate grammatical/thematic relations, especially in the sentences where
understanding semantics is not enough. Sentence 1 is not thematically ambiguous2; it is
obvious from the meaning of the words ‘tea’, ‘mother’, and ‘drink’ who the agent is even
though morphologically both nouns are unmarked.
(1) Əни чəй эчə
Əni çəy eçə
Ani çAi eçA
Mother tea drinks
‘Mother drinks tea’
In sentence 2, recognizing inflectional morphemes is important because we cannot
rely on semantics or word order to distinguish the agent from the patient.
(2) Малай-ны кыз ярата
Malay- nɪ kɪz yarata
Malay- nI kIz yarata
Boy-Acc girl-Nom love-Pres-P3-Sg
‘A/the girl loves the boy’
Knowing morphological processes can expand a language learner’s lexicon
acquisition and help in semantic, syntactic, and stylistic comprehension of a language
(Sabitova, 2002).
Morphological processing is a necessary and useful stage in NLP systems. In spell
checkers, for example, comparing an input word against a static list of words is impractical
and impossible due to language productivity: it is not possible to manually list all
morphological variants of every word3 in an agglutinating language in advance (Jurafsky &

2
Examples are listed in the following order: Cyrillic spelling, Latin spelling, xfst spelling (if different
from Latin spelling), gloss; Latin spelling and gloss are provided for a Tatar word mentioned in the discussion;
KaTMorph output examples are spelled with the alphabet used in the program (xfst column of Alphabet
conversion Table 1). The meanings of grammatical tags can be found in Table 3. Tatar language data, unless
specified otherwise, is self generated.
3
A Kazan Tatar noun has 52 unique forms, not counting derivational forms.
 4

Martin, 2008; Trost, 2003). However, a morphological analyzer, limited to mere word roots
with morphotactic rules, is the better alternative to the static word lists. For the purposes of
stemming, the process of stripping the word-end morphemes in Information Retrieval
systems, rule based morphological analyzers tend to be more accurate than traditional
stemmers (Tzoukerman, Klavans, & Strzalkowski, 2003). The morphological structure of a
word produced by an analyzer carries syntactic and grammatical information about the word,
information necessary for part of speech tagging and parsing sentences. This information is
especially valuable for dependency parsing, which is used for synthetic and agglutinating
languages.
Language choice for this project is determined by two factors: linguistic and
geopolitical. Tatar has rich but regular morphology accompanied by various phonologic
harmony rules, which makes it a suitable candidate for computing with finite-state networks.
Development of a morphological analyzer/generator requires implementation of
morphological knowledge and accompanying phonological processes, which creates a digital
record of a morphological model of a language. Despite its official status in the Republic of
Tatarstan, the number of native speakers of Tatar is declining. This tool has the potential to
contribute to language preservation and language learning as a stand-alone program or as a
stage in various NLP systems for Kazan Tatar in the future.
There are a few Turkic languages for which computational morphological analyzers
have been developed, including Turkish (Solak & Oflazer, 1992), Türkmen, and Crimean
Tatar (Altintas & Cicekli, 2001). The automated morphological analyzer is an integral part of
the Turkish spell-check application (Solak & Oflazer, 1992), the Turkish Dependency
Treebank Tagging Annotation tool (Atalay, Oflazer, & Say, 2003), and the machine
translation system between Turkish and Crimean Tatar (Altintas, 2001).
According to Suleymanov (2007), a two-level morphological analyzer was developed
for Kazan Tatar by the Research Lab at Tatarstan Academy of Sciences jointly with Bilkent
University; it is used for developing a machine translation system for the Turkish-Kazan
Tatar language pair. However, to my knowledge, it appears that the analyzer is available only
for the Kazan State University community.
 5

KAZAN TATAR LANGUAGE

4
Kazan Tatar belongs to the North-Western group of the Turkic languages
(Gadzhieva, 1990). It is one of the 3 main dialects of Tatar, the other two being Western
(Mishar), and Eastern (Siberian). Since 1992 it is one of the two official languages in the
Republic of Tatarstan, Russian Federation. In the Russian Federation, the term ‘Tatar’ refers
to Kazan Tatar. Throughout this discussion this term will also be in reference to Kazan Tatar.
According to the Russian census of 2002, there are over 5.5 million Tatars living in
the Russian Federation (Federal State Statistics Service, 2004), although it would be
inaccurate to assume that all are Tatar speakers. People that identify themselves of Tatar
ethnicity are not necessarily the native speakers of Tatar. According to the demographic
report by Tatarstan Academy of Sciences, demographic reasons such as urbanization, low
birth rate, intermixed marriages, and the fact that Tatar is not a prestige language, cause the
numbers of native speakers of Tatar to decline (Isxakov, 2007). To make matters worse, the
medium of instruction in the majority of Tatarstan educational institutions is Russian; so only
53% of Tatar children study in their native language (Svechnikov & Sergeeva, 2008).
Throughout history, the writing system of the Tatar language used several alphabets:
Arabic script was used until 1927, followed by Latin script until 1939 (Tatar language,
2009); currently the writing system is based on the Cyrillic alphabet. Alphabet
Transliteration Table 2 lists the Latin and Cyrillic alphabets along with the phonetic alphabet
which represents Kazan Tatar pronunciation. The ‘XFST’ column lists the modified alphabet
used in KaTMorph.
In Tatar, the basic word order is Subject Object Verb. However, non-subject NPs can
precede subjects for pragmatic reasons, such as expressing topicality. Adverbial phrases of
place and time can be sentence initial; nominal modifiers precede their heads. In literary
Tatar, the predicate is sentence final5 and can be expressed by a verb, noun, adjective, or
numeral. Verb and Agent agree in person and number. In copular sentences the copular verb
is present in the future and past tense, and obligatorily absent in the present tense. Sentence 3

4
Kazan is situated on the Volga River. Kazan Tatar is also referred to as Volga Tatar.
5
Prescriptive Tatar grammar prohibits verbs in non-final position. To determine whether verbs are always
sentence final, spoken language data needs to be gathered and examined.
 6

Table 2. Alphabet Transliteration

6
Latin Cyrillic XFST IPA
A A a ɑ
Ə Ə A æ
B Б b b
C Җ c ʑ
Ç Ч C tʃ
D Д d d
E E e e
F Ф f f
G Г g g,G
H H h h
I Ы I ɨ
İ И i i
J Ж j ʒ
K K k k,q
L Л l l
M M m m
N H n n
Ŋ Ң N ŋ
O O o o
Ɵ Ɵ O ø
P П p p
R P r r
S C s s
Ş Ш S ʃ
T T t t
U У u u
Ü Y U y
V B v v
X X x x
Y Й y j
Z З z z

6
Source for the “Latin” and “Cyrillic” columns is ‘Republic of Tatarstan’ (2010).
 7

has the canonical word order; and sentences 4, 5, and 6 are the examples of the clauses
without the copular.
(3) Син китап-ны укы-дың
sin kitap-nɪ ukɪ-dɪŋ
sin kitap-nI ukI-dIN.
You book-Sg-Def-Acc read-PastDef-P2-Sg
‘You read the book.’
(4) Бу укутучы
Bu ukutuçɪ
Bu ukutuCI
This teacher-Sg-Nom
‘This is a teacher.’
(5) Бу күлмəк иске
Bu külmək iske
Bu kUlmAk iske
This shirt-Sg-Nom old
‘This shirt is old.’
(6) Балалары өч-əү (adopted from Nasibullina, 2008)
Balalarɪ ɵç-əü
BalalarI Oç-AU
Child-PL-Poss-P3-Nom three-Pron-Coll
Their children three
‘They/She/He have/has three children’
Due to the subject-verb concord in person and number, subjects in a Tatar sentence
are optional. Both sentences 7 and 8 are grammatical; the subject in sentence 7 is inferred
from the verb’s suffix for person and number and is the same as in sentence 8. Even though
these two sentences are not pragmatically equivalent, they are semantically and thematically
equivalent.
 8

(7) Китапны укыдым

Kitap- nɪ ukɪ-dɪ-m
Kitap-nI ukI-dI-m
book-Acc-Def read-PastPerf-P1-Sg
‘I read the book’
(8) Мин китап-ны укыдым
Min kitap- nɪ ukɪ-dɪ-m
Min kitap- nI ukI-dI-m
I-Nom book-Acc-Def read-PastPerf-P1-Sg
‘I read the book’

Tatar Phonetic Model

The Tatar alphabet consists of 9 vowels and 22 consonants7. For the purpose of this
processor, vowels are subdivided into front and back vowels (in Tatar, corresponding
terminology is ‘soft’ and ‘hard’ vowels). Tables 3 and 4 illustrate transliteration between the
Latin vowels and the corresponding vowels used in the program. The vowels in the tables are
lined up with their corresponding pair. For example, the front ‘ü’ and the back ‘u’ form a
pair. It should be noted that in suffixes, the front vowel ‘e’ corresponds to the back vowel ‘ɪ’.
In addition, the vowel ‘ə’ is the orthographic representation of the sound [æ], and should not
be confused with the schwa sound.

Table 3. Front Vowels

Latin İ, i Ü, ü Ɵ, ɵ Ə, ə E, e
xfst i U O A e

Table 4. Back Vowels

I,
Latin ɪ U,u O,o A,a
xfst I u o a

7
Table 1 lists the Tatar alphabet transliteration.
 9

The Latin base spelling of the Tatar consonants and their xfst representation is
presented in Table 5.

Table 5. Tatar Consonants

Latin XFST Latin XFST
B, b b N, n n
C, c c Ŋ, ŋ N
Ç, ç C P, p p
D, d d R, r r
F, f f S, s s
G, g g Ş, ş S
H, h h T, t t
J, j j V, v v
K, k k X, x x
L, l l Y, y y
M, m m Z, z z

The main classification of the Tatar consonants relevant for the development of
KaTMorph is their division into voiced and voiceless consonants. The consonants used in the
processor and throughout this discussion are:
Voiced Consonants: [ b | c | d | v | g | z | j | m | n | N | y | l | r ]
Voiceless Consonants: [ p | C | t | f | k | s | S | h | x ]
In Tatar, every morpheme has several variants (allomorphs) which are phonologically
conditioned, with the exceptions of personal pronouns min ‘I’ and sin ‘you’, which are non-
harmonic in dative case. The maximum number of possible allomorphs per morpheme is six.
Allomorphs are derived from the underlying morpheme for the following reasons: (a) vowel
harmony; (b) nasal assimilation; (c) assimilation by the [voice] feature. The underlying form
(UF) is the starting point of the morphological derivation; the orthographic form after
applying phonological rules is the surface form (SF). It should be noted that the listed
phonological processes are not all-inclusive; they are selected based on their relevance for
 10

this project. These processes are chosen because they change the orthographic form of a
word.
(a) Vowel Harmony can be defined as the process of vowel assimilation to the preceding
vowel’s [back] feature. As a consequence, a Tatar word cannot contain both front and
back vowels, with the exceptions of foreign borrowings such as kitap ‘book’,
compound words such as ɵç-poçmak ‘three-corners’8, and previously mentioned
personal pronouns, which are non-harmonic. In borrowings and compounds the
[back] feature of a suffix vowel is chosen based on the vowel quality of the preceding
syllable. Compare the word pairs in Table 6 where ungrammatical words are marked
with an asterisk (*):

Table 6. Examples of Vowel Harmony Rules

Back Front Front-Back
Back bar-a 'goes' *bər-a kitap-lar 'books'
Front *bar-ə бəр-ə 'beats' *kitap-lər

(b) Consonant Nasal Assimilation in Tatar is progressive: consonants ‘l’ and ‘d’ copy the
[+nasal] feature of a preceding consonant. Examples 9 and 10 demonstrate how the
consonants in underlying forms of plural suffix –lar and ablative case suffix -dan
assimilate to the preceding nasal consonant and become –nar and –nan respectively.
(9) UF: tun-lar
SF: tun-nar
furcoat-PL-Nom
‘furcoats’
(10) UF: tun-dan
SF: tun-nan
furcoat-Sg-Abl
‘from furcoat’

8
çpoçmak ‘three-corners’ ≈ ‘triangle’ is the traditional Tatar pie made of meat and potatoes
 11

(c) Consonant Voice Assimilation results in a change of the [voice] feature which affects
the consonants of both bound and free morphemes. Example 11 illustrates how by
progressive assimilation, –d or –g at the beginning of the inflectional morphemes
become voiceless after voiceless consonants (Nasibullina, 2008).
(11) SF: kit-de kolak-ga
UF: kit-te kolak-ka
Leave-PastDef-P3-Sg ear-Sg-Dat
‘he/she left’ ‘to a/the ear’

The stem final consonants –p or –k become voiced in the intervocalic position, where
the right context vowel is a part of an inflectional morpheme (Nasibullina, 2008). Example
12 demonstrates that root-final voiceless consonants become voiced between vowels9:
(12) UF: kolak-ɪm tap-ar
SF: kolag-ɪm tab-ar
ear-Sg-Poss-P1-Sg find-FutIndef-P3-Sg
‘my ear’ ‘he/she will find’

Tatar Morphology
Morphology is the branch of linguistics that studies word structure. Morphological
knowledge is the knowledge of morphemes, and knowledge of the rules to combine them;
these rules are called morphotactic rules. A morpheme is the smallest meaningful unit of
language expressing semantic concepts or grammatical features (Fromkin, Rodman, &
Hyams, 2003). Morphological analysis is the process of identifying morphemes in a given
word. Morphemes that can be words are free morphemes, and the ones that must attach to
stems are bound morphemes. In Tatar, bound morphemes are added to a word by suffixing.
To form a new word with a new meaning a derivational affix is added. To change the

9
Other voiceless consonants, for example ‘t’, do not become voiced:
UF: kibet-em
SF: kibet-em
‘my store’
 12

grammatical meaning of a word inflectional morphemes are used; they don’t change the
syntactic category of words, but mark their properties such as case, gender, number, and
tense (Fromkin et al., 2003). Morphologically complex words consist of a free morpheme, or
root, which carries the main lexical content, and at least one derivational morpheme; thus a
new stem can be formed. By adding more affixes to the stem, yet more complex stems can be
formed (Fromkin et al., 2003). A bound morpheme can have several realizations which are in
complementary distribution to each other; they constitute a set of allomorphs.
As an agglutinative language, Tatar has rich and complex morphology. This richness
and complexity is defined by the number of the underlying suffixes, allomorphs, morpheme
functions, and recursive word formation devices. Tatar suffixes are used to derive new
words, to express grammatical relations or features, and to express pragmatic meaning
(Ganiev, 2000). There is an arsenal of over 300 derivational suffixes (Ganiev, 2000), and
word formation is recursive, so that a derived word can be a new stem for a new derivation.
Tatar bound morphemes are suffixes: they attach after a root or a stem. Each part of
speech has its own set of suffixes, and each suffix has a set of phonologically determined
allomorphs. The inflectional morphemes for nouns include case, number, and possessive
features; for verbs, they are tense, aspect, mood, person, number, and negation, just to name a
few (Ganiev, 2000). Grammatical details for each part of speech are provided in the
corresponding sections of this discussion.
It should be noted that there is one morpheme that can attach to any part of speech—it
is the ‘yes-no question’ morpheme. The polar question is formed by adding the underlying
interrogative suffix –mɪ to a predicate or to any word in a sentence for ‘contrastive purposes’
(Underhill, 1986, p.59). Examples 13 illustrate the question formation for major parts of
speech (for grammatical tag meanings, see Morphological Feature Tags, Table 7).
(13)
(a)
Син китап-ны укы-дың-мы?
sin kitap-nɪ ukɪ-dɪŋ-mɪ?
sin kitap-nI ukI-dIN-mI?
You book-Sg-Def-Acc read-PastDef-P2-Sg-Ques
‘Did you read the book?’
 13

Table 7. Morphological Feature Tags

Adj adjective
Noun noun
Verb verb
Pron pronoun
+ morpheme boundary
^DB-Adj adjective derived from another word
^DB-Noun noun derived from another word
P1 1st person
P2 2nd person
P3 3rd person
Sg singular
PL plural
Indef indefinite
Def definite
Nom nominative
Acc accusative
Gen genitive
Dat dative
Loc locative
Abl ablative
Abl2 alternative ablative
Poss possessive-personal
Coll collective
Pres present
PastIndef past indefinite
PastDef past definite
FutIndef future indefinite
FutDef future definite
Imp imperative
Neg negation
Ques interrogative
 14

(b)
Бу укутучы-мы?
Bu ukutuçɪ-mɪ?
Bu ukutuCI-mI?
This teacher-Sg-Nom-Ques
‘Is this a teacher?’
(c)
Бу күлмəк иске-ме?

Bu külmǝk iske-me?
Bu kUlmAk iske-me?
This shirt-Sg-Nom old-Ques
‘Is this shirt old?’
(d)
Мин-ме?
Min-me?
I-Ques
‘me?’
(e)
Бу-мы?
Bu-mɪ?
Bu-mI?
This-Ques
‘This?’
(f)
Китап-ны-мы?
Kitap-nɪ-mɪ?
Kitap-nI-mI?
Book-Acc-Def-Ques
‘The book?’(Obj)
 15

(g)
Барган-мы?
Bar-gan-mɪ?
Bar-gan-mI?
Go-PastIndef-P3-Sg-Ques
(he/she) went?
‘Did he/she go?’
(h)
Шат- сыз- лык- мы?
şat- sɪz- lɪk- mɪ?
Sat- sIz- lIk- mI?
Joyful-^DB-Adj-^DB-Noun-Ques
Joyful-without-N?
‘Unjoyfulness?’

MORPHOTACTIC RULES
The order of suffixes in a Tatar word is fixed. Tatar nouns, verbs, pronouns and
adjectives have their own set of morphotactic rules; however, these rules have the following
commonalities. Derivational suffixes immediately follow the root or another derivational
suffix, inflectional morphemes come next, and the interrogative morpheme is always word
final. Because word derivation is recursive, there could be more than one derivational suffix
in a word. The morphotactic rules used in the processor are summarized in Figure 1 (note
that derivational suffixes for verbs are not included in the current version of the program, but
listed here only for illustrative purpose).

Noun Root + Derivational Suffix + Plural Suffix + Personal Suffix + Case + Interrogative Suffix
Verb Root + Derivational Suffix + Negation + Tense + Personal Ending + Interrogative Suffix
Pronoun + Case + Interrogative Suffix
Adjective Root + Derivational Suffix + Interrogative Suffix
Figure 1. Morphotactic rules.
 16

RECURSIVE CONCATENATION
My present work has a focus on Tatar inflectional morphology with a goal to provide
morphological analysis of nouns, personal pronouns, adjectives, and verbs. However, I
included the minimal derivational morphology in the lexicon in order to demonstrate the
recursive behavior of word formation and to provide a framework for future development.
The noun lexicon has a finite number of morphemes: 12 Inflectional, and 8
derivational; yet the number of words that can be generated by the grammar is unlimited, as
is the word length. The grammar defined for nouns and adjectives is recursive, which allows
the generation of an infinite number of words. The circularity results from the loops in the
noun lexicon, specifically due to the derivational rules for noun and adjective formation.
Figure 2 is an illustration of the recursive behavior of the word generation from the root süz
‘word’ where new nouns and adjectives are formed by reapplying derivational morphemes to
the newly formed stems.

süz
süz-Noun+Sg+Nom
‘word’

süz-le
süz-Noun+^DB-Adj
‘talkative’ = verbose

süz-le-lek
süz-Noun+^DB-Adj+^DB-Noun+Sg+Nom
‘quality of being talkative’ = verbosity

süz-le-lek-sez
süz-Noun+^DB-Adj+^DB-Noun+^DB-Adj
‘the one without the quality of being talkative’ = the one without verbosity = taciturn (the
shorter version, süzsez, means ‘speechless’ )

süz-le-lek-sez-lek
süz-Noun+^DB-Adj+^DB-Noun+^DB-Adj+^DB-Noun+Sg+Nom
‘the quality of being without the quality of being talkative’ = taciturnity
Figure 2. Tatar word formation.

Even though these words are possible in the language, the comprehension of such
complex derived words gets more challenging as the word ‘grows’ longer. If necessary, it is
possible to restrict such recursive generation by applying a ‘filter’ on top of the lexicon
 17

(Beesley & Karttunen, 2003). This filter can eliminate the undesired strings based on the
limit we set on the number of unique derivational morphemes a word can have. The true
challenge is determining this limit. Another option is to set semantic restrictions, which
would allow only certain roots to be the stems for new derivations. To implement such
restriction, the most productive suffixes should be factored out from the set of the
derivational suffixes. This would allow all nouns or adjectives to get productive derivational
suffixes, and only a limited number of nouns or adjectives to get the non-productive ones10.
Note, however, that such restriction is not implemented in the current version of the
processor.
The following sections present an overview of morphological aspects of Tatar
language relevant to the design of this morphological processor.

NOUN
Grammatical categories of nouns are number, possession, and case; they are
expressed by inflectional suffixes. The underlying forms of inflectional and derivational
morphemes have multiple allomorphs conditioned by vowel harmony rules and nasal or
voice assimilation as discussed in the Tatar Phonetic Model section.
In a noun, suffixes are sequenced as follows: Noun Root + Derivational Suffix+
Inflectional Suffix + Interrogative Suffix. The inflectional morphemes for nouns are optional,
but they do have to attach in the following order: the plural suffix, a personal suffix, and a
case suffix. Thus, the morphological structure of a noun is schematically represented as
follows: Noun Root + Derivational Suffix + Plural Suffix + Personal Suffix + Case Suffix +
Interrogative Suffix.
The derivational suffixes form a new noun with new meaning by appending to noun
or adjective roots or stems. Table 8 demonstrates the most productive Tatar suffixes for noun
formation included in the KaTMorph lexicon (Nasibullina, 2008).
Due to the recursive noun formation, the maximum number of morphemes a noun can
have is unlimited, but the number of non-derivational suffixes is limited to four: plural,
personal, case, and question. Example 14 demonstrates how the noun is composed by the

10
See Beesley & Karttunen (2003, p.245) for details.
 18

Table 8. Noun Derivational Suffixes

stem suffix meaning example
noun ç person ‘doer’ süz-lek-çe ‘lexicographer’
noun, adj l k abstract noun süz-lek ‘dictionary’ ; şat-l k ‘joy’
noun l person from kazan-l ‘kazanien’

final noun rule. The noun is derived from an adjective and is in plural form functioning as a
direct object in a question form:
(14) Шат- лык- лар- ы- ны- мы
Sat- lIk- lar- I- nI- mI
Sat-Adj+^DB-Noun+ PL+ Poss+P3+ Def-Acc+ Ques
Joyous+Noun+PL+Poss+P3+Def+Acc+Ques
‘Their joys-Acc?’
Next, let me present an overview of Tatar noun grammar reflected in the lexicon of
the Morphological Processor.

Number
Tatar nominals have the grammatical feature of number. The plural form is marked
by the underlying suffix –lar attached to the noun root. This underlying form changes
according to the morphophonological rules. First, if the noun ends with the nasal consonant
[n, ŋ, m], then –l becomes –n. Second, -а- changes to -ə- when the base noun has a front
vowel in the final syllable. Table 9 demonstrates the surface forms of all 4 possible forms of
the plural morpheme.

Table 9. Tatar Noun Singular and Plural Forms

Sg Noun Singular Plural

[+back] vowels bala ‘child’ bala-lar ‘children’
[-back] vowels keşe ‘person’ keşe-lər ‘people’
nasal,[+back] vowels uram ‘street’ uram-nar ‘streets’
nasal, [+back] vowel kɵn ‘day’ kɵn-nər ‘days’
 19

It should be noted that the plurality feature can also be expressed syntactically, by
means of quantative pronouns or numerals with the noun in the unmarked form; the
processor will mark the nouns in such expressions as morphologically singular:
Күп бала
küp bala
kUp bala
Many child-Sg-Nom
‘many children’

Биш бала
Biş bala
Five child-Sg-Nom
‘Five children’

Personal Suffixes (Possessive)

There is a group of morphemes in Tatar indicating possession; there are 6 of them
distinguishing person and number of a possessor. These suffixes attach to the noun roots
referring to possessed thing(s) or person(s). These morphemes can attach either to a singular
or plural form of a noun. The surface forms of each morpheme are governed by phonological
processes discussed in the Phonetic Model Section. Table 10 lists the underlying personal
suffixes and Table 11 lists the declension of the noun ‘aunt’ by possession in nominative
case.

Table 10. Tatar Personal Suffixes

Sg Possessor post consonant or ü/u post vowels
1 ɪm m
2 ɪŋ ŋ
3 ɪ sɪ
PL Posessors
1 ɪbɪz bɪz
2 ɪgɪz gɪz
3 ɪ sɪ
 20

Table 11. Noun Declension by Possesion

Sg apa 'aunt' apa-lar 'aunts'
1 apa-m 'my aunt' apa-lar-ɪm 'my aunts'
2 apa-ŋ 'your aunt' apa-lar-ɪŋ 'your aunts'
3 apa-sɪ 'his/her aunt' apa-lar-ɪ 'his/her aunts'
PL
1 apa-bɪz 'our aunt' apa-lar-ɪbɪz 'our aunts'
2 apa-gɪz 'Your aunt' apa-lar-ɪgɪz 'Your aunts'
3 apa-sɪ 'their aunt' apa-lar-ɪ 'their aunts'

Case
Case is used to express grammatical relations between words in a sentence. The case
feature in Tatar is expressed syntactically by means of postpositions and morphologically by
means of suffixes. Current discussion is concerned with morphological cases only. Most
Tatar linguists agree that the Tatar noun has at least six morphological cases (Ganiev, 2000).
These cases are nominative, genitive, dative, accusative, ablative, and locative. Table 12 is a
declension table for Tatar nouns.

Table 12. Case Paradigm for Nouns

Case alma ‘apple’ et ‘dog’
Nominative alma et
Genitive alma-nɪŋ et-neŋ
Dative alma-ga et-kə
Accusative Definite alma-nɪ et-ne
Accusative Indefinite alma et
Ablative alma-dan et-tən
Locative alma-da et-tə

The dictionary form of a noun is listed in the morphologically unmarked, nominative

singular form. It has the functions of (a) the nominative case, (b) the indefinite accusative
case, and (c) the indefinite genitive case:
a) Nominative case
1) A noun functions as a subject of a sentence. In (15) the noun ‘woman’ is the
subject:
 21

(15) Бу апа бик матур.

Bu apa bik matur.
This woman very beautiful.
‘This woman is very beautiful’
2) A noun functions as a nominal predicate. In (16) ‘student’ is the nominal
predicate.
(16)Бу кыз студент
Bu kɪz student
Bu kIz student
this girl student
‘this girl is a student’
3) A noun functions as a vocative, as the noun iptəşlər ‘friends’ in sentence 17.
(17) Җырлыйк əле, иптəшлəр. (Tatar folk song)
сɪrlɪyk əle, iptəş-lər.
cIrlIyk Ale, iptAS-lAr.
Sing-Imp Imp-Part friend-Pl
‘Let us sing, friends’
b) Indefinite accusative case
A noun in this form functions as an unspecified or nonreferential direct object of a verb.
In (18) the noun ‘book’ is the morphologically unmarked direct object:
(18) Мин китап укыйм
Min kitap ukɪym
Min kitap ukIym
I -Nom book-Indef-Acc read-Pres-P1-Sg
‘I am reading a book’
c) Indefinite genitive case (Ganiev, 2000).
A noun in this form is nonreferential; ‘dog’ in (19) is such a noun.
(19) эт койрыг-ы
et koyrɪg-ɪ
et koyrIg-I
dog-Nom tail-Sg-Poss-P3-Sg
 22

dog’s tail
The accusative case marks a uniquely identifiable direct object of a verb by adding
the underlying suffix -nɪ. The direct object in (20) refers to the specific book.
(20) Мин китап-ны укыйм
Min kitap-nɪ ukɪym
Min kitap-nI ukIym
I book-Acc-Def read-Pres-P1-Sg
‘I am reading the book’
A noun in the dative case form has the underlying ending –ga and expresses time,
location, goal, or an indirect object of a verb (Zakiev & Ramazanova, 2002). In example (21)
the indirect object is marked with the dative case.
(21) Ул малай-га китап бирде
ul malai-ga kitap bir-de
He boy-Dat book-Acc-Indef give-PastDef-P3-Sg
‘He gave a boy a book’
The locative case, as the name suggests, expresses the location of the action or event:
ɵydə ‘at home’ (22). The same case is used to express the place in time: noyabrda ‘in
November’.
(22) Өй-дə ул миң-а китап бирде
ɵy-də ul miŋ-a kitap bir-de
Oy-dA ul miNa kitap bir-de
Home-Loc he I-Dat book-Acc-Indef give-PastDef-P3-Sg
‘At home, he gave me a book’
The literal translation of the ablative case from Tatar is a ‘point of departure’. It is
formed by adding the underlying –dan to the noun roots and it carries the functions of
expressing
(a) ‘the place from which’ or ‘the place through which’ (Lewis,1967)
өй-дəн
ɵy-dən
Oy-dAn
Home-Abl
 23

‘from home’
ишек-тəн

işek-tǝn
iSek-tAn
door-Abl
‘through door’
(b) ‘reason for a state’
шатлык-тан
şatlɪk-tan
SatlIk-tan
Joy-Abl
‘from (because of) joy’
(c) in comparative constructions
Казан Уруссу-дан зур-рак
Kazan Urussu-dan zur-rak
Kazan-Nom Urussu-Abl big-Comp
‘Kazan is bigger than Urussu’
(d) the material from which something is made
Бу алка-лар алтын-нан яса-л-ган
Bu alka-lar altɪn-nan yasa-l-gan
Bu alka-lar altIn-nan yasa-l-gan
This earring-PL gold-Abl make-Passive-PastIndef-P3-Sg
‘These earrings are made of gold’
The genitive case underlying morpheme -nɪŋ attaches to a noun referring to a
possessor (qualifying noun). It can attach to the root (23) or to the stem marked with personal
suffix (24).
(23) Апа-ның өй-е
аpa-nɪŋ ɵy-e
аpa-nIN Oy-e
aunt-Gen house-Sg-Poss-P3-Sg
 24

‘aunt’s house’
(24) Апа-лар-ыгыз- ның өй-е
аpa- lar- ɪgɪz- nɪŋ ɵy-e
apa- lar- IgIz- nIN Oy-e
aunt-PL- Poss-P2-PL-Gen house-Sg-Poss-P3-PL
aunts-yours-theirs house
‘your aunts’ house’

Definiteness
In Tatar, definiteness is expressed by morphological case alternation, so the NP in the
same grammatical function can have different case markings (Aissen, 1999). In Tatar,
Nominative-Accusative and Nominative-Genitive alternations are used to express the notions
of definiteness, referentiality, and specificity. Morphologically unmarked forms are for
indefinite or non-referential nouns, and the accusative and genitive suffixes are used to mark
definite nouns.
In canonical, nominative-accusative transitive constructions, the morphological
realization of the direct object case is dependent on definiteness: an indefinite direct object is
morphologically unmarked, and a definite direct object has the accusative suffix. Examples
25 and 26 illustrate the contrast in definiteness via direct object alternation: alma ‘apple’
functioning as the direct object is in the nominative form to express indefinite (25), and has
the accusative ending –nɪ to express the definite direct object (26).
(25) Ул алма ашады
ul alma-NULL aşadɪ
ul alma-NULL aSadI
She/He-Nom apple-NULL-Indef eat-PastDef-P3-Sg
‘She/He ate an apple11.’
‘She/He ate apples.’

11
Morphologically unmarked objects are number neutral (R, Underhill, personal communication,
November 2, 2010)
 25

(26) Ул алма-ны ашады

ul alma-nɪ aşadɪ
ul alma-nI aSadI
She/He-Nom apple-Acc-Def eat-PastDef-P3-Sg
‘He ate the apple.’
In possessive constructions, the modifying noun has a function of the genitive case;
its morphological realization is dependent on definiteness or referentiality of the possessor to
which the noun is referring. NP constructions in 27 and 28 are examples of the Genitive-
Nominative alternation. In (27) ‘dog’ is a generic, nonreferential noun, and ‘mother’ is a
definite noun (28).
(27) эт койры-гы
et koyrɪg-ɪ
et koyrIg-I
dog-Nom tail-Sg-Poss-P3-Sg
‘dog’s tail’
(28) əни-нең сумкас-ы
əni-neŋ sumka-sɪ
Ani-neN sumka-sI
mother-Gen purse-Poss-P3-Sg
‘mother’s purse’
In NP (27) the phrase-final head noun is marked with the possessive suffix of the 3rd
person and the modifying nonreferential noun is in nominative form. These types of
constructions are best described as Noun-Noun compounds. The relationship between the
two nouns is not possessive, but merely ‘qualificatory’ (Lewis, 1967). In NP (28) the head
noun is marked with the possessive suffix in the 3rd person, and a referential possessor has
the genitive marker (Nasibullina, 2008).
In addition to case marking, direct object definiteness can be expressed by determiners.
1) With definite nouns marked by the overt accusative ending
a. Optional determiner bu ‘this/these’ can be used:
bu alma-nɪ
this apple-Acc
 26

‘this apple’(Obj)
b. Optional determiner ber ‘one’ is used to express the quantity:
Ber alma-nɪ
one apple-Def-Acc
‘one (particular) apple’(Obj)
c. Optional possessive pronoun can be used:
Min-em xatɪn-nɪ
my wife-Def-Acc12
‘my wife’(Obj)
2) With indefinite nouns without case marking the optional determiner ber ‘one/a’ can
be used:
ber alma
one apple-Indef-Acc
‘one/an apple’(Obj)
It should be noted that KaTMorph will return two analyses for the noun in the
unmarked form. For example, there are two parses for the word eş ‘work’:
eS-Noun+Sg+Nom
eS-Noun+Sg+Indef-Acc
Nouns marked by personal suffixes are definite by definition; therefore they cannot
be analyzed as indefinite accusative. This processor will produce an unambiguous analysis
for such forms:
eş-ebez
eS-Noun+Sg+Poss+P1+PL+Nom
‘our work’
To summarize, Tatar noun paradigms defined in the processor’s lexicon include
(structure is adopted from Underhill (1986)):
1. Noun Root
2. Derivational Suffixes

12
the same meaning ‘my wife-Acc’ can be expressed by means of a personal morpheme -ɪm: xatɪn-ɪm-
nɪ. Yet, one more way to express the same semantic meaning is minem xatɪn-ɪm-nɪ.
 27

3. Plural
4. Personal Suffixes
5. Case
The inflectional paradigm for the word apa ‘aunt’ is presented in the series of Table
13a-d. Note that there are no accusative indefinite forms for the nouns marked with
possessive suffixes.

Table 13a. Declension of apa

aunt aunts
Sg PL
Nom apa apalar
Gen apanɪŋ apalarnɪŋ
Dat apaga apalarga
Acc-Def apanɪ apalarnɪ
Acc-Indef apa apalar
Abl apadan apalardan
Loc apada apalarda

Table 13b. Declension of apa, 1st Person Possessor

my aunt my aunts our aunt our aunts
Sg Poss P1 Sg PL Poss P1 Sg Sg Poss P1 PL PL Poss P1 PL
Nom apam apalarɪm apabɪz apalarɪbɪz
Gen apamnɪŋ apalarɪmnɪŋ apabɪznɪŋ apalarɪbɪznɪŋ
Dat apama apalarɪma apabɪzga apalarɪbɪzga
Acc-Def apamnɪ apalarɪmnɪ apabɪznɪ apalarɪbɪznɪ
Acc-Indef -- -- -- --
Abl apamnan apalarɪmnan apabɪzdan apalarɪbɪzdan
Loc apamda apalarɪmda apabɪzda apalarɪbɪzda

Table 13c. Declension of apa, 2nd Person Possessor

your-Sg aunt your-Sg aunts Your-Pl aunt Your-Pl aunts
Sg Poss P2 Sg PL Poss P2 Sg Sg Poss P2 PL PL Poss P2 PL
Nom apaŋ apalarɪŋ apagɪz apalarɪgɪz
Gen apaŋɪŋ apalarɪŋnɪŋ apagɪznɪŋ apalarɪgɪznɪŋ
Dat apaŋa apalarɪŋa apagɪzga apalarɪgɪzga
Acc-Def apaŋnɪ apalarɪŋnɪ apagɪznɪ apalarɪgɪznɪ
Acc-Indef -- -- -- --
Abl apaŋnan apalarɪŋnan apagɪzdan apalarɪgɪzdan
Loc apaŋda apalarɪŋda apagɪzda apalarɪgɪzda
 28

Table 13d. Declension of apa, 3rd Person Possessor

his/her/their aunt his/her/their aunts
Sg Poss P3 PL Poss P3
Nom apasɪ apalarɪ
Genitive apasɪnɪŋ apalarɪnɪŋ
Dat apasɪna apalarɪna
Acc-Def apasɪn apalarɪn
Acc-Indef -- --
Abl apasɪnnan apalarɪnnan
Loc apasɪnda apalarɪnda

VERB
Next, let me present an overview of Tatar verb morphology relevant to this project.
The dictionary form of a verb is Infinitive, for example, bararga ‘to go’. The root of a verb is
its imperative 2nd person singular form such as bar ‘go’. It is verb roots, not dictionary forms,
that are lexical entries in this morphological processor (the reasoning for this choice is
discussed in the Lexicon section of the Methodology Section).
Verb grammatical features defined in this program are aspect, tense, person, number
and negation (Ganiev, 2000); they are expressed via morphology. By the morphotactic rules,
an optional derivational morpheme13 can be added to the stem, followed by an optional
negation morpheme, a tense morpheme, a personal suffix, and ending with an optional
interrogative suffix; schematically these rules are represented as follows: Verb Root +
Derivational Suffix + Negation + Tense + Personal Ending + Interrogative Suffix.
Verb suffixes representing its grammatical features will be discussed in the remaining part of
this section and they are:
1. Tense: Present, Past Indefinite, Past Definite, Future Indefinite, Future Definite
2. Personal: Person and Number

Negation
To negate a sentence with a verbal predicate, the negation morpheme is added after
the verb root before the tense or mood suffix. Examples 29 and 30 illustrate the negated
sentences in the past definite tense and the imperative mood.

13
The verb derivational morphology is not implemented in the current version of this program.
 29

(29) Син китапны укы-ма-дың

sin kitapnɪ ukɪ-ma-dɪŋ
sin kitapnI ukI-ma-dIN
You book-Sg-Def-Acc read-not-PastDef-P2-Sg
‘You did not read the book’
(30) Кил- мə- сен
Kil- mə- sen
Kil- mA- sen
Come-not- Imp-P3-Sg
‘Let him/her not come’
The negation suffix has 3 allomorphs: -ma, -mə, and -m. Verbs in the future and past
tenses and the imperative mood get –ma/-mə14 and the present tense verbs get the suffix -m
(Nasibullina, 2008). Table 14 illustrates the negation allomorph distribution for the 3rd person
Singular.

Table 14. Verb Negation

Imperative Present Past Indef Future Indef
‘come' kil-mə-sen kil-m-i kil-mə-gən kil-mə-s
‘go' bar-ma-sɪn bar-m-ɪy bar-ma-gan bar-ma-s

A verb in the negative form has the same tense morphemes as its positive counterpart,
with the exception of the present and future indefinite tense. Table 15 has both affirmative
and negated verb forms for bar ‘go’ conjugated by tense for the 3rd person singular.

Table 15. Verb Conjugation by Tense for the 3rd

Person Singular
Affirmative Negative
Imperative bar-sɪn bar-ma-sɪn
Present bar-a bar-m-ɪy
Past Indef bar-gan bar-ma-gan
Past Def bar-dɪ bar-ma-dɪ
Fut Indef bar-ɪr bar-ma-s
Fut Def bar-açak bar-ma-yaçak
14
with the vowel choice determined by the vowel harmony rules
 30

Mood
Currently, KaTMorph can process verbs in the imperative and indicative moods15.
Verbs in the imperative mood denote order, request, or invitation and conjugate by person
and number for the 2nd and 3rd Person. The conjugation for the verb kilergə ‘to come’ in the
imperative mood can be seen in Table 16.

Table 16. Imperative Mood Conjugation

Person/Number Singular Plural

1 -- --
2 kil kil-egez
3 kil-sen kil-sennər

Note that the translations for the 3rd person forms are equivalent to the English ‘Let
him/her/it/them verb’ or ‘Make her/him/it/them verb’. Sentence (31) exemplifies the use of
the verb in the imperative mood.
(31) Бəхет килсə, бүген кил-сен (tatar folk song)
Bəxet kilsə, bügen kil-sen
bAxet kilsA, bUgen kil-sen
happiness come-Cond, today come-Imp-P3-Sg
‘If happiness comes, let it come today’
Tatar verb conjugation includes adding tense or mood morphemes, followed by
personal endings. There are two types of the personal endings. Type I is used after the past
indefinite and future tenses; Type II is for the past definite tense. The present tense personal
endings are of Type I for plural forms and of mixed type for singular. Table 17 lists the
allomorphs of the personal suffixes for both types. Full verb paradigm for each tense
mentioned will be presented in the corresponding sections of the Tense Morphology section.

15
The subjunctive and conditional mood suffixes can be added in the future version of the program after
more research has been completed.
 31

Table 17. Verb Personal Suffixes

Type I Type II
Singular back front back front
P1 mɪn men m m
P2 sɪŋ seŋ ŋ ŋ
P3 NULL NULL NULL NULL
Plural
P1 bɪz bez k k
P2 sɪz sez gɪz gez
P3 lar lər lar lər

Tense
Tatar verbs have three tenses with the dimension of definetness: past, present, and
future. The verb referring to the action happening at the moment of speaking is in the present
tense, before the moment of speaking in the past tense, and after the speech event in the
future tense.
When affirmative, a verb in the present tense has –a, -ɪy, or –i suffix attached to the
root, followed by personal suffixes of Type I for plural and of mixed type for singular:
1. The root ending with a consonant gets suffix –a: bar-a ‘goes’, kil-ə ‘comes’. Some of
the exceptions are yɵr-i ‘walks’, avɪr-ɪy ‘hurts’.
2. The root ending with a back vowel gets suffix –ɪy and the root final vowel gets
deleted. Observe how the surface form differs from the ungrammatical UF:
UF: ukɪ-ɪy
SF: uk-ɪy
‘reads’
3. The root ending with a front vowel gets suffix –i and the root final vowel gets
deleted:
UF: tɵze-i
SF: tɵz-i
‘builds’
For negated verb forms in the present tense suffixes -ɪy/-i are used: bar-m-ɪy ‘he
does not go’, kil-m-i ‘he does not come’, ukɪ-m-ɪy ‘she does not read’, tɵze-m-i ‘she does not
 32

build’. The present tense paradigm for both affirmative and negated forms of the verb tap
‘find’ is presented in Table 18.

Table 18. Present Tense Paradigm

affirmative negative
Person Singular Plural Singular Plural
tap-m-ɪy-
1 tab-a-m tab-a-bɪz tap-m-ɪy-m bɪz
tap-m-ɪy- tap-m-ɪy-
2 tab-a-sɪŋ tab-a-sɪz sɪŋ sɪz
3 tab-a tab-a-lar tap-m-ɪy tap-m-ɪy-lar

The past tense has two aspects with their own inflections: indefinite and definite. The
past indefinite is used to express an event that the speakers themselves did not observe or
cannot remember well; it is commonly used for describing historical facts or in narratives
(Nasibullina, 2008). The past indefinite is formed by adding the underlying suffix –gan to the
root, followed by personal inflections of Type I (Nasibullina, 2008). Past definite refers to
the event the reality of which is without any doubt (Nasibullina, 2008) and has the underlying
morpheme –de after the root, followed by personal suffixes of Type II (see Table 19 for the
past tense paradigm). Sentences 31 and 32 illustrate the difference between the definite and
indefinite aspects of the past tense.
(31) Ул кичə кинога бар-ган.
Ul kiçə kinoga bar-gan
Ul kiCA kinoga bar-gan
He-Nom yesterday movie-Dat go-PastIndef-P3-Sg
‘Yesterday, he supposedly went to the movies.’
(32) Ул кичə кинога бар-ды.
Ul kiçə kinoga bar-dɪ
Ul kiCA kinoga bar-dI
He-Nom yesterday movie-Dat go-PastDef-P3-Sg
‘Yesterday, he went to the movies.’ (The speaker either witnessed this event or is
absolutely certain it happened)
 33

Table 19. Verb Conjugation for the Past Tense

Indefinite Definite
Person Sg PL Sg PL
1 tap-kan-mɪn tap-kan-bɪz tap-tɪ-m tap-tɪ-k
2 tap-kan-sɪŋ tap-kan-sɪz tap-tɪ-ŋ tap-tɪ-gɪz
3 tap-kan tap-kan-nar tap-tɪ tap-tɪ-lar

Like the past tense, the future has indefinite and definite aspects. The definite is used
when the future event will certainly happen (Nasibullina, 2008). In order to form the future
definite, the underlying morpheme –аçak is added to the root; to form the future indefinite,
the underlying suffix -ar is added to the root, followed by the personal suffixes of Type I.
Table 20 lists the future tense forms of the verb ‘find’.

Table 20. Verb Conjugation for the Future Tense

Indefinite Definite
Person Sg PL Sg PL
1 tab-ar-mɪn tab-ar-bɪz tab-açak-mɪn tab-açak-bɪz
2 tab-ar-sɪŋ tab-ar-sɪz tab-açak-sɪŋ tab-açak-sɪz
3 tab-ar tab-ar-lar tab-açak tab-açak-lar

To summarize, Tatar verb paradigms defined in the processor’s lexicon include

(structure adopted from Underhill (1986)):
1. Verb Root
2. Negation
3. Tense
4. Personal Suffixes
Finally, example 33 demonstrates a parsed verb with all possible morphemes:
(33) Бар-ма- ды- лар- мы
bar-ma- dɪ- lar- mɪ
bar-ma- dI- lar- mI
go- Neg-PastDef- P3-PL-Ques
go-not-Past-they-?
‘did they not go?’
The complete verb paradigm for tap ‘find’ is presented in Appendix A.
 34

PRONOUNS
Tatar pronouns are subdivided into personal, demonstrative, interrogative, relative,
indefinite, and possessive (Nasibullina, 2008). The current version of the morphological
processor can parse or generate only personal pronouns. Adding the rest of the pronouns to
the system is a matter of adding their stems to the lexicon. The following information about
Tatar personal pronouns is included in the lexicon.
In Tatar, there are three persons for pronouns: the 1st is to denote the speaker, the 2nd-
for person being addressed, and the 3rd- for anybody else. The pronouns are devoid of
gender; the third person singular pronoun ul is gender neutral. The pronouns have plural
forms and the 2nd Plural form coincides with the 2nd Formal form. In addition, Tatar personal
pronoun paradigm includes declension by six morphological cases16. It should be noted that
there are some irregularities in pronoun morphology. First, the pronouns min ‘I’ and sin ‘you’
are non-harmonic in dative case. Second, the paradigm of the pronoun ul ‘he/she/it’ has
irregular morphological structure: even though the non-nominative forms have the
phonologically consistent root an-, this root is not the dictionary form of the 3rd person
singular pronoun, which is ul. Table 21 represents Tatar personal pronoun paradigm.

Table 21. Personal Pronouns Paradigm

Case P1-Sg P2-Sg P3-Sg P1-PL P2-PL P3-PL
Nom min ‘I’ sin ‘you’ ul ‘he/she/it’ bez ‘we’ sez ‘you’ alar ‘they’
Gen minem sineŋ anɪŋ bezneŋ sezneŋ alarnɪŋ
Dat miŋa siŋa aŋa bezgə sezgə alarga
Acc mine sine anɪ bezne sezne alarnɪ
Abl minnən sinnən annan, aŋardan bezdən sezdən alardan
Loc mində sində anda, aŋarda bezdə sezdə alarda

If a sentence consists of only a personal pronoun, to form a polar question the

underlying interrogative suffix –mɪ is added at the end of the word. The sequence of pronoun
morphemes is Root + Case + Interrogative. Examples (34) illustrate pronoun morphotactics.
(34) (а) Мин- ме?

16
The interrogative pronouns ‘who’ and ‘what’ have the case feature as well, and can be added to the
pronominal lexicon in the future.
 35

Min- me?
I –Nom- Ques
Me? (‘who, me?’)
(b) Син-ең- ме?
sin- eŋ- me?
Sin- eN- me
You-Gen-Ques
‘Yours?’
(с) Алар-да- мы?
alar- da- mɪ?
alar- da- mI
They-Loc-Ques?
At them?
‘at their place?’

ADJECTIVE
The main purpose of the current morphological processor is to analyze and produce
noun, personal pronoun, and verb paradigms. Adjectives were added to the program in order
to demonstrate word formation in Tatar. New nouns can be derived from adjectives by
adding derivational suffixes, and likewise, new adjectives can be formed by adding suffixes
to nouns. The derivational suffix is appended to the noun root or stem. Table 22 demonstrates
the underlying forms of the most productive suffixes for adjective formation, which are
included in the lexicon (Nasibullina, 2008).

Table 22. Adjective Derivational Morphemes

stem suffix example
noun lɪ bolɪt-lɪ ‘cloudy’
noun sɪz təm-sez ‘tasteless’
noun çan eş-çǝn ‘hard-working’

The interrogative suffix can be added to the end of adjectives if this adjective is
contrastive or is a nominal predicate. This is the only inflectional suffix that can be handled
 36

for adjectives in the current version of the processor. Examples (35) show parses for
adjectives.
(35) (a) матур
matur
matur-Adj
‘beautiful’
(b) матур- мы?
matur- mɪ?
matur- mI?
matur-Adj+Ques
‘(is he/she/it) beautiful?’

METHODOLOGY
For the past thirty years most work in computational morphology has been heavily
dependent on finite-state methods, with the dominating approach based on finite-state
transducers. Douglas Johnson (1972) was the first to demonstrate that phonological rewrite
rules describe regular relations and, theoretically, can be implemented as finite-state
transducers. Around 1980, Xerox researchers from Palo Alto Kaplan and Kay rediscovered
this idea (Kaplan & Kay, 1994). Later, the Xerox Palo Alto Research Center developed
algorithms for finite-state computing (Karttunen & Beesley, 2005). Since then, large-scale
implementations of morphology for most European languages, Turkish, Arabic, Korean, and
Japanese based on finite-state tools were developed at Xerox (Karttunen, 2001). Xerox finite-
state tools are language independent tools that linguists can use to create finite-state
transducers for various NLP tasks. These tools include lexc, a high-level language “to specify
natural language lexicons” (Beesley & Karttunen, 2003, p. xv), and xfst, which provides an
interface with regular expression compiler and access to the “algorithms of the Finite-State
Calculus” (Beesley & Karttunen, 2003, p. 81).
A finite-state morphological analyzer and generator, which can be called a Lexical
Transducer, is a two-sided network: the upper-side represents lexical strings and the lower-
side represents surface strings. The lexical level string consists of the elements of the word’s
paradigm: root and tags representing grammatical features of the word. The surface string is
 37

a word as it is represented in the original language (Altintas, 2001). The Lexical Transducer
is bidirectional: in an analysis, the transducer relates a lower string to an upper string; in a
generation the input string is applied to the upper side of the transducer, and if it is in the
upper language the transducer will return the surface form(s) (Beesley & Karttunen, 2003).
The goal of the Lexical Transducer is to address the two main components of
morphology: morphotactics and morphophonological alternation (Beesley & Karttunen,
2003). Morphotactic rules are encoded as finite-state networks in a finite-state lexicon. The
finite-state lexicon, generally developed by a linguist, specifies roots, affixes, and
morphotactics. It generates morphotactically well-formed strings (Beesley & Karttunen,
2003); the traditional linguistic name for such a form is an underlying form (UF).
Phonological alternations co-occurring with morphological processes are implemented by
replace rules as finite-state transducers17. These rules are usually designed by a linguist as
well. Rule Transducers map an underlying form to the properly spelled surface string or form
(SF) (Beesley & Karttunen, 2003). To sum up, the lexicon network, composed together with
the replace rules, allows bidirectional mapping between the upper and lower strings.

Implementation
KaTMorph is implemented using Xerox’s proprietary finite-state tools and
techniques.
This system is divided into lexicon and rule modules. When lexicons and rules are
composed, the resulting sublanguages are unioned together into a single lexical transducer
which can process nouns, personal pronouns, adjectives, and verbs. In KaTMorph, lexicons
are separate for verbs and nominals, each with its own set of replace rules which affect
orthography. Lexicons and rules are based on the morphological and phonological processes
discussed in the Kazan Tatar Language Section.
This processor works bidirectionally: it maps grammatical words to lexical forms, and
vice versa. Figure 3 shows the architecture of such morphology system (Beesley &
Karttunen, 2003).

17
The xfst replace rules are traditionally called rewrite rules in linguistics.
 38

Upper Language
Root+Suff1+Suff2
↕
Lexc Grammar
↕
Intermediate Language
↕
Replace Rule Grammar
↕
Lower language
‘surface string’
Figure 3. FST architecture.

A lexical string contains a word root with a sequence of morphemes ordered

according to the morphotactic rules. These rules are explained in detail in the Tatar
Morphology section. Derivational morphemes are marked by symbols such as ^DB-POS;
they indicate that the root’s part of speech has changed or the same part of speech with a new
meaning is formed at this boundary. In order to decipher the lexical form, consult Table 7
Morphological Feature Tags on page 13 which lists the tags and their meaning used in this
program. The tags for the lexical language were designed in consideration with the existing
NLP systems for Turkic languages, such as the METU-Sabanci Turkish Treebank (Oflazer,
Say, Hakkani-Tür, & Tür, 2003) and the Crimean Tatar morphological analyzer (Altintas &
Cicekli, 2001).
Since the mapping between strings is bidirectional, it allows both production of valid
words from the upper string, and analysis of a word from the lower string. When the word
barmassɪŋ ‘you will not go’ is analyzed, the returned solution has the lexical string
consisting of the root and grammatical tags: bar-Verb +Neg +FutIndef +P2-Sg which reads
as follows:
1. The root is ‘bar’
2. Part of speech is verb
3. It is a negated form
 39

4. Future indefinite tense

5. 2nd Person, Singular
When a word is generated from the root and a sequence of tags, such as bar-
Verb+Neg+FutIndef+P2-Sg, each symbol is matched with the upper side symbols, and if the
match is successful, the surface string barmassɪŋ ‘you will not go’ will be returned.
Appendix B demonstrates the sample runs of the morphological analysis for noun paradigm
apa ‘aunt’ produced by KaTMorph. The sample runs of the verb generation from the root bar
‘go’ is presented in Appendix B.
The lexicons and rules for KaTMorph were guided by Nasibullina ‘Tatar language in
Tables and Schemes’ (2008). The choice of word roots was determined by the examples
discussed by Nasibullina (2008) and by a frequency dictionary I compiled from a literary
Tatar corpus18. Besides the root choices, the word list served as the testing data for
morphological analysis during the testing and debugging phases of the development.
There are a total of 53 roots, 37 inflectional morphemes, and 8 derivational
morphemes in the KaTMorph lexicon. The alphabet used in the program is Latin based with
some modifications. Table 1 demonstrates the correspondence between the xfst alphabet,
used in this program, and the Latin or Cyrillic alphabet used in Tatar orthography. The
current version of the processor can only be executed in the environment where the xfst
software has been installed19. The complete source code is presented in Appendix C. Let us
now discuss in detail the lexicon and the replace rules implemented in KaTMorph.

Lexicon
The lexicon compiler (lexc) is a XEROX formalism used to describe finite-state
networks (Beesley & Karttunen, 2003). The lexc is used to define Tatar morphotactics: word
roots, suffixes, and rules for combining them. The lower-side language of the lexc transducer
is an intermediate representation of a word which needs to be mapped to grammatical surface
strings via the replace rules.

18
I have constructed a corpus of Tatar literary language consisting of over 1,123,000 words from the
publically available Tatar Electronic Library (http://kitap.net.ru). The text was randomly selected from over 20
authors dating from 1912 to the present time.
19
link to the xfst web site: www.fsmbook.com
 40

There are two separate lexicons: for verbs and for nominals. In order to keep the
system “concise, maintainable and easily expandable” (Beesley & Karttunen, 2003, p. 264),
the following considerations from Beesley & Karttunen (2003) were taken into account while
developing lexicons:
1. To have a minimally necessary set of unique morphemes, only the underlying forms
were entered. Regular morphophonological alternations to accommodate allomorphy
are handled by replace rules. At the same time, there is a need to handle irregular
strings in the lexc to avoid deriving them via the replace rules.
2. To design lexicons in a way that reduces the future development to the simple
addition of new morphemes by a lexicographer without prior knowledge of xfst.
3. To group morphemes by grammatical features for ease of maintaining the project and
better readability and accessibility.
The lexc file consists of several lexicons with unique names. The entries inside a
lexicon follow these templates: ‘Form Continuation Class’ or ‘upper:lower Continuation
Class’. The Form field is a string of characters representing morphemes; ‘upper:lower’
format indicates a string of the upper language to be mapped to the lower; Continuation Class
is the name of another lexicon within the same file to which the given form can ‘continue’ to
form stems (Beesley & Karttunen, 2003).
The lexc baseforms, referred to as roots throughout the discussion, were chosen based
on the regularities of word formation or affixation for each part of speech. The baseforms for
a noun, personal pronoun, and adjective are dictionary forms; for verbs, it is the stem, which
is in the imperative 2nd person singular form. These choices are dictated by the goal to keep
the replace rules as simple as possible. For example, if I chose the verb baseforms to be the
infinitive, dictionary form, such as bararga ‘to go’, then the infinitive marker for each verb
derivation or affixation would have to be deleted. However, the inflectional and derivational
suffixes attach to the verb in the imperative 2nd person singular form, hence the choice of the
baseform.
Morphemes are grouped into the lexicons by the features and functions they share;
morphotactic rules are implemented via continuation classes (Beesley & Karttunen, 2003).
The continuation classes for nouns are derivational and inflectional suffixes; the former
derive nouns and adjectives, the latter include plural, case, and personal morphemes. Verbs
continuation classes include tenses, the imperative mood, personal suffixes, and negation.
 41

Both nouns and verbs have the ability to form a yes-no question form with the Interrogative
continuation class, which finalizes the word formation.
Figure 4 shows the lexicons for adjectives. The entries in the LEXICON Adj indicate
A as their continuation class. The LEXICON A consists of three types of suffixes: Interog,
DerivNoun2, and DerivAdj2, each has its own continuation class. The continuation class for
DerivAdj2 is AdjSuff, which means that after the derivational suffix –sɪz has been added,
any of the three suffixes, Interog, DerivNoun2 and DerivAdj2, can follow. The continuation
class pound sign (‘#’) in the LEXICON Interog indicates the end of a word: this means that
no morphemes can attach after the question morpheme –mɪ. Optionality of a morpheme is
expressed by including an empty entry in the same LEXICON. Thus, the question morpheme
is optional. Another way to express optionality is creating an intermediate LEXICON, such
as the LEXICON AdjSuff in Figure 4, which makes all adjective morphemes optional
(Beesley & Karttunen, 2003). The recursive word formation is expressed via loops. An
example of such a loop is in the DerivAdj2 lexicon. In this segment of the lexc file, the
continuation class of DerivAdj2 is AdjSuff, the LEXICON in which DerivAdj2 itself resides
(Beesley & Karttunen, 2003).

Morphophonological Rules
The xfst replace rules are what linguists call phonological rewrite rules (Beesley &
Karttunen, 2003). The basic conditional replace rule has the following template A -> B || L _
R where A, B, L, and R are arbitrarily complex regular expressions denoting regular
languages. L and R represent the left and right context for the rule to fire. This replace rule
describes a relation that maps the upper string A to the lower string B if A is between L and
R (Karttunen, 1995). After the compilation of the morphotactic rules to a finite-state
transducer, they are joined with the replace rules which transform the lexical forms into the
surface forms.
Recall that each inflectional and derivational Tatar morpheme has multiple
allomorphs conditioned by phonological processes. To keep the morphotactic rules in lexc
file as general as possible, the goal is to limit the number of morphemes for each feature
strictly to the underlying form. It is the replace rules that can do the work of modifying the
underlying forms to the surface forms based on the phonological environment. All of the
 42

LEXICON Adj
matur A ; !beautiful
iske A ; !old
Sat A ; !joyful

LEXICON A
-Adj:0 AdjSuff;

LEXICON AdjSuff
Interog ;
DerivNoun2 ; ! optionally attach to adjective to form a noun
DerivAdj2 ; ! optional suffix after adjective stem to form an adjective

LEXICON DerivNoun2
+^DB-Noun:lIk N ; !this is a recursive loop: it will take a new noun and will add
!deriv suffixes;

LEXICON DerivAdj2
+^DB-Adj:sIz AdjSuff ; !after adj, form adjective then a noun

LEXICON Interog
+Ques:mI # ;
#; !Optional interogatory –mI

Figure 4. Adjective lexicon.

underlying forms for suffixes in this processor have been arbitrarily chosen to have the
[+voice, -nasal] features for the initial consonant, and the [+back] feature for vowels. Thus,
the underlying form for the ablative case, for example, is –dan, and all possible allomorphs
for the ablative case are: -dan, -dən, -tan,-tən, -nan, and -nən.
The replace rules in this processor are motivated by regular phonological processes
existing in Tatar such as vowel harmony, assimilation, epenthesis, and deletion, and by some
idiosyncrasies in word formation. They were adopted from Nasibullina (2008) and Ganiev
(2000) and discussed in the Phonetic Model Section. The processes of epenthesis and
deletion of segments, a vowel or a consonant, are to prevent a syllable from violating the
syllable structure principles of a language20 (Spencer, 1996). Tatar syllable structures

20
Segment deletion and insertion are used to avoid vowel hiatus at the morpheme boundary (R.Underhill,
 43

followed by examples are presented below (Nasibullina, 2008), where V is for Vowel, and C
is for Consonant:
V: I ‘bend’
VC: ak ‘white’
CV: su ‘water’
CVC: kiŋ ‘wide’
VCC: ant ‘oath’
CVCC: kart ‘old’
The replace rules which derive grammatical word forms from lexical forms in
KaTMorph are outlined in this section. These rules are presented in the order in which they
apply in the derivation. It should be noted that these replace rules do not apply to word roots,
with the exceptions of root-final u/ü, rule 6, and root-final voiceless consonants, rule 7. To
restrict these rules from firing within a root, the suffix boundaries are marked in the
intermediate strings with the symbols “+” for the morpheme boundary21 (R. Malouf, personal
communication, October 25, 2010; R. Underhill, personal communication, November 2,
2010). These rules are the generalized versions of the replace rules used in the program. For
literal representations of these rules see Appendix C, Source Code in files tatar-noun-
rule.regex and tatar-verb-rule.regex and see Table 23 (Karttunen, 1995) for regular
expression notation.
1. Nasal Assimilation
If a noun root ends with a nasal consonant [n, ŋ, m], then –l in the plural morpheme
and –d in the ablative morpheme become –n:
[ l | d ] -> n || [m|n|N] + _
UF: urman-lar urman-dan
SF: urman-nar urman-nan
‘forests’ ‘from forest’
2. –g deletion after N or m

personal communication, November 2,2010)

21
Note, that the morpheme boundary symbol will not be seen by the users.
 44

Table 23. Regular Expressions Notation

Symbol Meaning
.#. word boundary symbol
[..] empty string
| OR, the union operator
+ concatenation one or more times
* Kleene-Star: concatenation zero or more times
() optionality
[] grouping brackets
.o. composition
.x. cross-product
-> replacement
_ site of replacement
|| replace rule and context separator

This rule deletes –g in the dative suffix, when preceding -N or -m is a part of the
personal morpheme. However, this rule does not delete -g if the nasal consonant is a part of a
root because the presence of the morphemes ‘Im’ or ‘In’ is required in the left context:
g -> [..] || + I [m | N ] + _
UF: apa-ɪŋ-ga apa-ɪm-ga taŋ-ga kiem-gə
IF: apa-ɪŋ-a apa-ɪm-a taŋ-ga kiem-gə
SF22: apa-ŋ-a apa-m-a taŋ-ga kiem-gə
‘to your aunt’ ‘to my aunt’ ‘to sunrise’ ‘to clothes’
3. -I deletion
This rule applies to the words marked with the personal possessive suffixes which
start with the underlying vowel -ɪ. The vowel -ɪ is dropped if a noun root ends with a
vowel23, except u/ü:
+ I -> [..] || [eIiaAo] _
UF: apa-ɪm su-ɪm baş-ɪm kino-ɪm
SF: apa-m su-ɪm baş-ɪm kino-m
‘my aunt’ ‘my water’ ‘my head’ ‘my movie’

22
The Surface Form is derived after the application of rule 3.
23
The vowels o/ø never occur in the final syllable in Tatar words (Nasibullina, 2008), except for the
borrowings, such as the Russian borrowing kino ‘movie’.
 45

4.–t epenthesis with the personal morphemes

This rule inserts -t if a noun root ends with –s and an underlying personal morpheme
which starts with ‘ɪ’ is present. It is possible that dus ‘friend’ is an irregular noun. Then its
paradigm should be explicitly entered to the lexicon and not derived via morphotactic and
replace rules. However, at this time I was not able to find examples of other nouns with –s
final roots to check for similar behavior. This issue will be addressed in the future version of
the program after more research of language data has been done.
[..] -> t || s _ + ɪ
UF: dus-ɪm
SF: dus-t-ɪm
‘my friend’
5. –s deletion in the personal morphemes
If a noun root ends with a consonant or a diphthong (ending with u/ü/y), -s in the
underlying 3rd Person personal morpheme is deleted:
s I -> I || [ Cons | Vowel [ u | ü ] ] + _ ]
UF: baş-sɪ sorau-sɪ apa-sɪ
IF24: baş-ɪ sorau-ɪ apa-sɪ
SF: baş-ɪ sorav-ɪ apa-sɪ
‘his head’ ‘her question’ ‘his aunt’
6. u and ü become v
Root final –u or -ü become –v if it is between vowels, where the right context vowel
is the beginning of an inflectional morpheme:
[u|ü] -> v || Vowel _ + I
SF: sorau-ɪm ülçəü-ɪm
UF: sorav-ɪm ülçəv-em
‘my question’ ‘my heel’
7. Voice assimilation

24
The IF for sorau is ungrammatical; the SF is derived with rule 6.
 46

Root-final voiceless consonants become voiced in the intervocalic position, where the
right context vowel is the beginning of an underlying bound morpheme. Most inflectional
morphemes start with -ɪ, except the future definite and the present tense morphemes which
begin with -a:
k -> g, p -> b || Vowel _ + [ I | a ]
UF: kolak-ɪm tap-a tap-a
SF: kolag-ɪm tab-a tab-aCak
‘my ear’ ‘finds’ ‘he/she will find’
Inflectional morphemes starting with a consonant assimilate to the root final
consonant voice feature. Since the underlying forms were chosen to have a voiced initial
consonant, the rule devoices these consonants when preceded by a voiceless consonant.
g -> k, d -> t || VoicelessCons + _
UF: kolak-da kit-gan
SF: kolak-ta kit-kən
‘to an/the ear’ ‘departed’
8. Vowel harmony
Due to the fact that the underlying forms of suffixes were chosen to contain a back
vowel, these replace rules target the words where the root has a front vowel in the final
syllable. Rule (a) maps the suffix vowels a to A and I to e, if the root final syllable has a front
vowel, as in the word keşe ‘person’. Recall that the borrowings and noun-compounds are
non-harmonic, yet the rule (a) will not apply to the vowels within such roots because of the
morpheme boundary restriction.
(a)
a -> A, I -> e || FrontVowel (Cons)* + (Cons) _
The rule (b) ensures that the final syllable of polysyllabic suffixes, such as –IbIz for the
possessive 1st person plural, undergoes the transformation as well.
(b)
a -> A, I -> e || + (Cons) FrontVowel (Cons)+ _
 47

UF: keşe-lar keşe-sɪz keşe-lar-ɪbɪz kitap-lar25

SF: keşe-lAr keşe-sez keşe-lAr-ebez kitap-lar
‘people’ ‘without a person’ ‘our people’ ‘books’
In addition to non-harmonic borrowings, other exceptions to the vowel harmony rules
are personal pronouns in dative case siŋa ‘to you’ and miŋa ‘to me’. These exceptions are
entered into the lexicon in their final surface forms. For example, the dative form of ‘I’ is not
derived from the root with the dative suffix: the surface string miNa is explicitly mapped to
the lexical string min-Pron+Dat. Because these surface strings are not derived in the dative
case, there are no explicit morpheme boundaries “+”, so the vowel harmony rules do not
apply in these forms. However, the question forms of these words: siŋa-mɪ ‘to you?’ and
miŋa-mɪ ‘to me ?’ are derived from the dative stems by the morphotactic rules. That allows
the vowel in the interrogative suffix to remain [+back].
9.–y with the personal morphemes
When the personal morphemes are added to the stems which end with –y-final
diphthongs, -y is deleted and -ɪ/i is converted to –e:
y + [I|i] -> e || Vowel _
UF: abɪy-ɪm ətkəy-ɪm
SF: abɪ-em ətkə-em
‘my uncle’ ‘my father’
10. –y in negation
The present tense suffix in the negated form drops its final –y if the root final vowel
is fronted. The optionality of the consonant in the left context is justified by the need to apply
this rule to the words with CVV structure (bie ‘dance’):
y -> [..] || FrontVowel (Cons)* + m + i _
UF: bar-mɪy kil-mɪy bie-mɪy kiy-mɪy
SF: bar-mɪy kil-mi bie-mi kiy-mi
‘does not go’ ‘does not come’ ‘does not dance’ ‘does not put on’

25
kitap ‘book’ is an Arabic borrowing
 48

CONCLUSION AND FUTURE RESEARCH

Applications for NLP such as machine translation, information retrieval, and spell
checkers may benefit from the syntactic analysis as a part of their system. Syntactic relations
of agglutinative languages, like Turkic languages, are expressed by affixation; therefore,
understanding morphology of such languages is crucial in constructing syntactic analysis.
Finite-state methods for language processing are efficient and practical, especially for
morphology of agglutinative languages because in such languages morphemes systematically
encode morphosyntactic features in linear orders (Roark & Sproat, 2007).
The goal of this project is to create a tool, KaTMorph, which allows the exploration
of Tatar language morphology. The tool was developed using Xerox finite-state toolkit. The
main sources of descriptive generalizations used in this project are Nasibullina (2008) and
Lewis (1967). The lexicon contains a finite-state description of Kazan Tatar pronominal and
verb morphology, which includes morpheme classes, morphotactic and morphophonological
rules.
The limitation of the rule-based morphological processors is the inability to handle
words with unknown morphemes: if a root or a bound morpheme is not in the lexicon
morphological analysis or production fails. This system can be expanded to handle more
roots and grammatical features. To expand the roots lexicon, words from the frequency
dictionary can be added to the system. A supplementary solution to this challenge is to
implement a Guesser which can be used to parse or produce words with stems or morphemes
even if they are not included in the lexicon (Beesley & Karttunen, 2003).
Finite-state approach for morphology is not equipped to resolve ambiguity because
the context in which a word form is found is not considered in the morphological analysis.
Therefore, the problem of syncretism, common for agglutinative languages26 can not be
resolved by this approach. Likewise, part of speech ambiguity, exemplified by the pair of

homonyms eş-lǝr (the noun ‘work’ in the plural form) and eşlǝ-r (the verb ‘to work’ in the
future indefinite tense), can not be resolved by the processor27.

26
For example, in Tatar, as discussed in the Noun section, the dictionary form of a noun has the functions
of the nominative, indefinite accusative and indefinite genitive.
27
The parses returned by KaTMorph for the form eSlAr are
 49

The recursive nature of Tatar word formation results in recursive loops which allow
an unlimited number of words and words of unlimited length. Even though recursive
production is not harmful for morphological analysis, it might not be desirable for
production. This is because well-formed words should occur in the language and be able to
be comprehended by native speakers. To restrict such generation, we can limit the number of
unique derivational morphemes a word can have by applying a ‘filter’ on top of the
overgenerating lexicon (Beesley & Karttunen, 2003). The true challenge is determining this
limit. This task can be supplemented by a Tatar corpus study and by investigating the
cognitive processes involved in the usage and processing of such words. Another option is to
set semantic restrictions, which would allow only certain roots to be the stems for new
derivations.
To utilize the computing power of transducers, translation pairs of Tatar word stems
and a language of choice, can be added in future development of this tool. Similarly,
transliteration of the xfst alphabet to Cyrillic or Latin can also be implemented.
KaTMorph is a prototype of the computational morphological analyzer/generator for
Kazan Tatar. In its current state it is useful for linguists who wish to understand the
morphological processes of Tatar, as well as for language learners to aid in their language
comprehension and the practice of word conjugation or declension . When KaTMorph
contains a complete description of Tatar morphology, it will be a useful tool for large-scale
NLP applications in the future.

eS-Noun+PL+Nom
eS-Noun+PL+Indef-Acc
eSlA-Verb+FutIndef+P3-Sg
 50

REFERENCES

Aissen, J. (1999). Markedness and subject choice in optimality theory. Natural Language &
Linguistic Theory, 17(4), 673-711.
Altintas, K.(2001). Turkish to Crimean Tatar machine translation system (Master’s thesis,
Bilkent University). Retrieved from http://www.cs.bilkent.edu.tr/
~ilyas/PDF/tainn2001-morph.pdf
Altintas, K., & Cicekli, I. (2001). A morphological analyzer for Crimean Tatar. Retreived
from http://www.cs.bilkent.edu.tr/~ilyas/PDF/tainn2001-morph.pdf
Atalay,N., Oflazer, K., & Say, B.(2003) The annotation process in the Turkish treebank. In
Proceedings of the EACL Workshop on Linguistically Interpreted Corpora-LINC,
Budapest, Hungary, April 13-14.
Beesley K. R., & Karttunen L. (2003). Finite state morphology. Stanford, CA: CSLI.
Federal State Statistics Service. (2004). All Russia population census 2002. Retrieved from
http://www.perepis2002.ru/index.html?id=17
Fromkin, V., Rodman R., & Hyams, N. (2003). An introduction to language. Boston, MA:
Heinle & Thomson.
Gadzhieva, N. Z. (1990). Turkic languages. Retrieved from http://www.philology.ru/
linguistics4/gadzhiyeva-90.htm
Ganiev, F. A. (2000). Tatar language: Problems and research. Kazan, Russia: Tatar Book
Publishing.
Isxakov, D. (2007). Tatars before 2025: Demographic report. Retrieved from
http://tatpolit.ru/category/ zvezda/2007-10-05/474#comment-5688
Johnson, C. D. (1972). Formal aspects of phonological description. The Hague, Netherlands:
Mouton Publishers.
Jurafsky, D., & Martin, J. H. (2008). Speech and language processing: An introduction to
Natural Language Processing, computational linguistics, and speech recognition.
Upper Saddle River, NJ: Pearson Education.
Kaplan, R. M., & Kay, M. (1994). Regular models of phonological rule systems. Retrieved
from http://www.aclweb.org/anthology/J/J94/J94-3001.pdf
Karttunen, L. (1995). The replace operator. Proceedings of the 33rd annual meeting on
Association for Computational Linguistics, Cambridge, MA, 16-23.
doi:10.3115/981658.981661
Karttunen, L. (2001). Applications of finite-state transducers in Natural Language
Processing. In S. Yu & A. Paun (Eds.), Implementation and application of automata
(pp. 34-46). Heidelberg, Germany: Springer Verlag.
 51

Karttunen, L., & Beesley, K. R. (2005). Twenty-five years of finite-state morphology.

Retrieved from http://csli-publications.stanford.edu/koskenniemi-festschrift/8-
karttunen-beesley.pdf
Lewis, G. L. (1967). Turkish grammar. Oxford, UK: Oxford University Press.
Nasibullina, R. N. (2008). Tatar language in tables and schemes for Russian speaking
students of elementary school. Kazan, Russia: Gyilem Publisher.
Oflazer,K., Say, B., Hakkani-Tür, D., & Tür,G. (2003). Building a Turkish treebank. In A.
Abeille (Ed.), Building and exploiting syntactically-annotated corpora (in press).
Boston, MA: Kluwer Academic Publishers.
Republic of Tatarstan. (2010). Tatar alphabet based on the Latin script. Retrieved from
http://www.tatar.ru/append200_a.html
Roark, B., & Sproat, R. (2007). Computational approaches to morphology and syntax. New
York, NY: Oxford University Press.
Sabitova, I.I. (2002). Description of word formation of Tatar lexicon. Kazan, Russia: Fiker.
Solak, A., & Oflazer, K. (1992). Parsing agglutinative word structures and its application to
spelling checking for Turkish. In Proceedings of COLING’92—14th Conference on
Computational Linguistics, Nantes, France, July 1992. Retrieved from
http://ebookpedia.net/PARSING-AGGLUTINATIVE-WORD-STRUCTURES-AND-
ITS-APPLICATION-TO----.html
Spencer, A. (1996). Phonology: Theory and description. Oxford, UK: Blackwell Publishing.
Suleymanov, Dj. (2007). Tatar language and infromation technology. Kazan, Russia:
Kazanskiy Federalist.
Svechnikov, A., & Sergeeva, M. (2008). Preservation of linguistic diversity: Russian
experience. Retrieved from http://www.ifapcom.ru/files/publications/sb_eng.pdf
Tatar language. (2009).Online Encyclopedia Krugosvet. Retrieved from http://www.
krugosvet.ru/enc/gumanitarnye_nauki/lingvistika/TATARSKI_YAZIK.html
Trost, H.(2003). Morphology. In R. Mitkov (Ed.), The Oxford handbook of computational
linguistics (pp. 25-47). Oxford, UK: Oxford University Press.
Tzoukerman, E., Klavans, J. L., & Strzalkowski, T. (2003). Information retrieval. In R.
Mitkov (Ed.), The Oxford handbook of computational linguistics (pp.529-544).
Oxford, England: Oxford University Press.
Underhill, R. (1986). Turkish. In D. I. Slobin & K.Zimmer (Eds.), Studies in Turkish
linguistics, (pp.7-23). Amsterdam, Netherlands: John Benjamins Publishing Co.
Zakiev, M. Z., & Ramazanova, D. B. (Eds.). (2002). Current questions of Tatar philology.
Kazan, Russian: Fiker.
 52

APPENDIX A

TATAR VERB PARADIGM

Table 24. Verb Paradigm of Tabarga ‘to Find’

Imperative Present Past Indef Past Def Fututre Indef Future Def
aff neg aff neg aff neg aff neg aff neg aff neg
Base tap tap-ma tab-a tap-mɪy tap-kan tap-ma-gan tap-tɪ tap-ma-dɪ tab-ar tap-ma-s tab-açak tap-ma-yaçak
Singular mix mix type1 type1 type2 type2 type1 mix type1 type1
P1 tab-a-m tap-mɪy-m tap-kan-mɪn tap-ma-gan-mɪn tap-tɪ-m tap-ma-dɪ-m tab-ar-mɪn tap-ma-m tab-açak-mɪn tap-ma-yaçak-mɪn
P2 tap tap-ma taba-sɪŋ tap-mɪy-sɪŋ tap-kan-sɪŋ tap-ma-gan-sɪŋ tap-tɪ-ŋ tap-ma-dɪ-ŋ tab-ar-sɪŋ tap-ma-s-sɪŋ tab-açak-sɪŋ tap-ma-yaçak-sɪŋ
P3 tap-sɪn tap-ma-sɪn tab-a tap-mɪy tap-kan tap-ma-gan tap-tɪ tap-ma-dɪ tab-ar tap-ma-s tab-açak tap-ma-yaçak
Plural type1 type1 type1 type1 type2 type2 type1 type1 type1 type1
P1 -- -- taba-bɪz tap-mɪy-bɪz tap-kan-bɪz tap-ma-gan-bɪz tap-tɪk tap-ma-dɪ-k tab-ar-bɪz tap-ma-bɪz tab-açak-bɪz tap-ma-yaçak-bɪz
P2 tab-ɪgɪz tap-ma-gɪz taba-sɪz tap-mɪy-sɪz tap-kan-sɪz tap-ma-gan-sɪz tap-tɪ-gɪz tap-ma-dɪ-gɪz tab-ar-sɪz tap-ma-s-sɪz tab-açak-sɪz tap-ma-yaçak-sɪz
P3 tap-sɪnnar tap-ma-sɪnnar taba-lar tap-mɪy-lar tap-kan-nar tap-ma-gan-nar tap-tɪ-lar tap-ma-dɪ-lar tab-ar-lar tap-ma-s-lar tab-açak-lar tap-ma-yaçak-lar
53
 54

APPENDIX B

EXAMPLES OF MORPHOLOGICAL ANALYSIS

AND GENERATION
Sample runs of morphological analysis of the noun inflectional paradigm apa ‘aunt’

apa apa-Noun+PL+Poss+P1+PL+Def-Acc
apa-Noun+Sg+Indef-Acc apamnan
apa-Noun+Sg+Nom apa-Noun+Sg+Poss+P1+Sg+Abl apalarIbIzga
apa-Noun+PL+Poss+P1+PL+Dat
apanI apamda
apa-Noun+Sg+Def-Acc apa-Noun+Sg+Poss+P1+Sg+Loc apalarIbIzdan
apa-Noun+PL+Poss+P1+PL+Abl
apaga apalarIm
apa-Noun+Sg+Dat apa-Noun+PL+Poss+P1+Sg+Nom apalarIbIzda
apa-Noun+PL+Poss+P1+PL+Loc
apadan apalarImnI
apa-Noun+Sg+Abl apa-Noun+PL+Poss+P1+Sg+Def-Acc apaN
apa-Noun+Sg+Poss+P2+Sg+Nom
apada apalarIma
apa-Noun+Sg+Loc apa-Noun+PL+Poss+P1+Sg+Dat apaNnI
apa-Noun+Sg+Poss+P2+Sg+Def-Acc
apalar apalarImnan
apa-Noun+PL+Indef-Acc apa-Noun+PL+Poss+P1+Sg+Abl apaNa
apa-Noun+PL+Nom apa-Noun+Sg+Poss+P2+Sg+Dat
apalarImda
apalarnI apa-Noun+PL+Poss+P1+Sg+Loc apaNnan
apa-Noun+PL+Def-Acc apa-Noun+Sg+Poss+P2+Sg+Abl
apabIz
apalarga apa-Noun+Sg+Poss+P1+PL+Nom apaNda
apa-Noun+PL+Dat apa-Noun+Sg+Poss+P2+Sg+Loc
apabIznI
apalardan apa-Noun+Sg+Poss+P1+PL+Def-Acc apalarIN
apa-Noun+PL+Abl apa-Noun+PL+Poss+P2+Sg+Nom
apabIzga
apalarda apa-Noun+Sg+Poss+P1+PL+Dat apalarINnI
apa-Noun+PL+Loc apa-Noun+PL+Poss+P2+Sg+Def-Acc
apabIzdan
apam apa-Noun+Sg+Poss+P1+PL+Abl apalarINa
apa-Noun+Sg+Poss+P1+Sg+Nom apa-Noun+PL+Poss+P2+Sg+Dat
apabIzda
apa-Noun+Sg+Poss+P1+PL+Loc apalarINnan
apamnI apa-Noun+PL+Poss+P2+Sg+Abl
apa-Noun+Sg+Poss+P1+Sg+Def-Acc apalarIbIz
apa-Noun+PL+Poss+P1+PL+Nom apalarINda
apama apa-Noun+PL+Poss+P2+Sg+Loc
apa-Noun+Sg+Poss+P1+Sg+Dat apalarIbIznI apagIz
55
apa-Noun+Sg+Poss+P2+PL+Nom apalarI
apa-Noun+PL+Poss+P3+Nom
apagIznI
apa-Noun+Sg+Poss+P2+PL+Def-Acc apalarIn
apa-Noun+PL+Poss+P3+Def-Acc
apagIzga
apa-Noun+Sg+Poss+P2+PL+Dat apalarIna
apa-Noun+PL+Poss+P3+Dat
apagIzdan
apa-Noun+Sg+Poss+P2+PL+Abl apalarInnan
apa-Noun+PL+Poss+P3+Abl
apagIzda
apa-Noun+Sg+Poss+P2+PL+Loc apalarInda
apa-Noun+PL+Poss+P3+Loc
apalarIgIz
apa-Noun+PL+Poss+P2+PL+Nom

apalarIgIznI
apa-Noun+PL+Poss+P2+PL+Def-Acc

apalarIgIzga
apa-Noun+PL+Poss+P2+PL+Dat

apalarIgIzdan
apa-Noun+PL+Poss+P2+PL+Abl

apalarIgIzda
apa-Noun+PL+Poss+P2+PL+Loc

apasI
apa-Noun+Sg+Poss+P3+Nom

apasIn
apa-Noun+Sg+Poss+P3+Def-Acc

apasIna
apa-Noun+Sg+Poss+P3+Dat

apasInnan
apa-Noun+Sg+Poss+P3+Abl

apasInda
apa-Noun+Sg+Poss+P3+Loc
56
Sample Runs of Generation of the Verb Paradigm Bararga ‘To Go’

bar-Verb+Pres+P1-Sg bar-Verb+PastIndef+P2-Sg
baram bargansIN bar-Verb+FutIndef+P3-Sg
barIr
bar-Verb+Pres+P2-Sg
barasIN bar-Verb+PastIndef+P3-Sg bar-Verb+FutIndef+P1-PL
bargan barIrbIz
bar-Verb+Pres+P3-Sg
bara bar-Verb+PastIndef+P1-PL
barganbIz bar-Verb+FutIndef+P2-PL
bar-Verb+Pres+P1-PL barIrsIz
barabIz bar-Verb+PastIndef+P2-PL
bargansIz bar-Verb+FutIndef+P3-PL
bar-Verb+Pres+P2-PL barIrlar
barasIz bar-Verb+PastIndef+P3-PL
bargannar bar-Verb+Imp-P2-Sg
bar-Verb+Pres+P3-PL bar
baralar bar-Verb+FutDef+P1-Sg
baraCakmIn bar-Verb+Imp+P3-Sg
bar-Verb+PastDef+P1-Sg barsIn
bardIm bar-Verb+FutDef+P2-Sg
baraCaksIN bar-Verb+Imp+P2-PL
bar-Verb+PastDef+P2-Sg barIgIz
bardIN bar-Verb+FutDef+P3-Sg
baraCak bar-Verb+Imp+P3-PL
bar-Verb+PastDef+P3-Sg barsInnar
bardI bar-Verb+FutDef+P1-PL
baraCakbIz bar-Verb+Pres+P1-Sg
bar+PastDef+P1-PL baram
bardIk bar-Verb+FutDef+P2-PL
baraCaksIz
bar-Verb+PastDef+P2-PL bar-Verb+Neg+FutIndef+P1-Sg
bardIgIz bar-Verb+FutDef+P3-PL barmam
baraCaklar
bar-Verb+PastDef+P3-PL bar-Verb+Neg+PastDef+P1-Sg
bardIlar bar-Verb+FutIndef+P1-Sg barmadIm
barIrmIn
bar-Verb+PastIndef+P1-Sg bar-Verb+Neg+PastDef+P1-Sg+Ques
barganmIn bar-Verb+FutIndef+P2-Sg barmadImmI
barIrsIN
57
 58

APPENDIX C

KATMORPH XFST IMPLEMENTATION

Source Code of KaTMorph
#tatar-script.xfst
#xfst script that reads and compiles the lexicon and rule files
#and composes the rule network under the lexicon network
#2010, Albina Davliyeva

#Upper Language
# |
#Lexc Grammar
# |
#Intermediate language
# |
# Rule Grammar
# |
#Lower Language

clear stack

define FrontVowel [ i U O A e ] ;
define BackVowel [ I u o a ] ;
define Vowel [i U O A e I u o a ] ;

define Cons [bdgGjcClmnNrvzqpfktsSxhw] ;

define VoicedCons [bcdgGjcylmnNrvz] ;
define VoicelessCons [pfkqtsSC] ;

# read replace rules for nouns

read regex < tatar-noun-rule.regex
define Rules

# read noun lexicon

read lexc < tatar-noun-lex.txt
define LexiconN

# read replace rules for verbs

read regex < tatar-verb-rule.regex
define Rules1

# read verb lexicon

read lexc < tatar-verb-lex.txt
define LexiconV

#Compose rules
read regex [ [ LexiconN .o. Rules ] [ LexiconV .o. Rules1 ] ] ;

save stack tatar.fst

59
! tatar-noun-lex.txt
! This file defines noun, pronoun, adjective roots and suffixes and morphotactics
! 2010

! Do not declare suffixes as multi-character symbol, as they would be interpreted as a single character,
! and no replace rule would work

Multichar_Symbols
-Noun -Pron -Adj
+Sg +PL -Indef -Def
+Nom +Acc +Poss +Gen +Dat +Loc +Abl +Abl2
+P1 +P2 +P3
+Ques
+^DB-Noun
+^DB-Adj

LEXICON Root
Noun ;
Pronoun ;
Adj ;
Number ;
Poss ;
PossPL ;
Case ;
CasePron ;
PossPron ;
N ;
A ;
Pr ;
Interog ;
NSuff ;
NSuff2 ;
NSuff3 ;
Deriv ;
DerivNoun ;
DerivNoun2 ;
DerivAdj ;
DerivAdj2 ;

!!!!!
! Lexical Entries by part of speech
!!!!!

!!!!!
! Nouns: noun stems are dictionary forms of nouns
!!!!!
60
LEXICON Noun
Ani N ; !mother
abIy N ; !uncle
alma N ; !apple
avIl N ; !village
apa N ; !aunt, woman
atna N ; !week
baS N ; !head
bolIt N ; !clowd
dus N ; !friend
eS N ; !work
kibet N ; !store
kiem N; !clothes
kolak N ; !ear
kOzge N ; !mirror
mAktAp N ;!school
Oy N ; !house,home
OstAl N; !table
uku N ; !study
UlCAU N ; !heel
UlAn N ; !grass
urIndIk N ;!stool
uram N ; !street
urman N ; !forest
sorau N; !question
sUz N ; !word
sUzlek N ; !dictionary
taN N; !sunset
yul N ; !road
kitap N ; !book (an Arabic borrowing, non-harmonic)

LEXICON Pronoun
min Pr;
sin Pr;
bez Pr;
alar Pr;
sez Pr;

LEXICON Pr
-Pron:0 CasePron ;

! Irregular forms of personal pronouns

LEXICON PossPron
min-Pron+P1+Sg+Poss:minem Interog ;
sin-Pron+P2+Sg+Poss:sineN Interog ;
ul-Pron+P3+Sg+Poss:anIN Interog ;
61
bez-Pron+P1+Pl+Poss:bezneN Interog ;
sez-Pron+P2+Pl+Poss:sezneN Interog ;
alar-Pron+P3+Pl+Poss:alarnIN Interog ;
min-Pron+Dat:miNa Interog ;
sin-Pron+Dat:siNa Interog ;
bez-Pron+Dat:bezgA Interog ;
sez-Pron+Dat:sezgA Interog ;
alar-Pron+Dat:alarga Interog ;
ul-Pron+Nom:ul Interog ;
ul-Pron+Acc+Def:anI Interog ;
ul-Pron+Dat:aNa Interog ;
ul-Pron+Abl:annan Interog ;
ul-Pron+Abl2:aNardan Interog ;
ul-Pron+Loc:anda Interog ;
ul-Pron+Loc2:aNarda Interog ;

LEXICON Adj
matur A ; !beautiful
iske A ; !old
Sat A ; !joyful

!!!
! Bound Morphemes are defined here
! The suffixes start with '7' to mark the suffix boundary to restrict the replace rules to apply only on these boundaries
!!!

!this will output part of speech tag for nouns

LEXICON N
-Noun:0 NSuff ;

!Noun root is followed by Plural or Derivational suffix

LEXICON NSuff
Number ;
Deriv ;

LEXICON Deriv
DerivNoun ;
DerivAdj ;

!suffixes to derive new nouns from noun stems

LEXICON DerivNoun
+^DB-Noun:7CI Number ;
+^DB-Noun:7daS Number ;
+^DB-Noun:7lIk NSuff ; !recursive loop

!suffixes to form new adjectives from noun stems

LEXICON DerivAdj
+^DB-Adj:7sIz AdjSuff ; !after noun , form adjective: dus-sIz then noun again: dus-sIz-lIk-mI
62
+^DB-Adj:7lI AdjSuff ;
+^DB-Adj:7Can AdjSuff ;
+^DB-Adj:7dagI AdjSuff ;

!number affix is followed by Possessive affix or by Null ending for Nom and Acc-Indef
LEXICON Number
+PL:7lar NSuff2 ;
+Sg:0 NSuff3 ;

LEXICON NSuff2
PossPL ;
Acc-Indef ; !Indef_acc case cannot follow Possessive affix

LEXICON NSuff3
Poss ;
Acc-Indef ; !Indef_acc case cannot followe Possessive affix

LEXICON PossCommon
+Poss+P1+Sg:7Im Case ;
+Poss+P1+PL:7IbIz Case ;
+Poss+P2+Sg:7IN Case ;
+Poss+P2+PL:7IgIz Case ;

LEXICON Poss
PossCommon ;
+Poss+P3:7sI Case ; ! the same form for Sg or Pl possessors, but only after Sg Noun
Case ; !Possessive is optional

LEXICON PossPL
PossCommon;
+Poss+P3:7I Case ; ! the same form for Sg or Pl possessors, but only after PL Noun
Case ; !Possessive is optional

LEXICON Case
Nom ; !Null ending for Nominative case
Gen ;
Acc-Def ;
Dat ;
Loc ;
Abl ;

LEXICON CasePron
Nom ; !Null ending for Nominative case
Acc-Def ;
Loc ;
Abl ;
63
LEXICON Nom
+Nom:0 Interog ;

LEXICON Acc-Def
+Def-Acc:7nI Interog ;

LEXICON Acc-Indef
+Indef-Acc:0 Interog ; !indefinite nouns get no case ending in accusative form

LEXICON Dat
+Dat:7ga Interog ;

LEXICON Loc
+Loc:7da Interog ;

LEXICON Abl
+Abl:7dan Interog ;

LEXICON Gen
+Gen:7nIN Interog ;

LEXICON Interog
+Ques:7mI # ; ! Interrogative suffix is the last suffix in a word
# ; ! Optional interrogatory -mI

LEXICON A
-Adj:0 AdjSuff;

LEXICON AdjSuff
Interog ;
DerivNoun2 ; !this will attach to adjective to form a noun
DerivAdj2 ; ! after adjective stem to form adjective -without

LEXICON DerivNoun2
+^DB-Noun:7lIk NSuff ; !this is recursive loop: will take a new noun and will add deriv suffixes; change to Number to avoid the
infinite loop

LEXICON DerivAdj2
+^DB-Adj:7sIz AdjSuff ; !after adj , form adjective: Sat-sIz then noun again: dus-sIz-lIk-sIz ; also recursive loop

!tatar-verb-lex.txt
!Verb morphology is defined here.

Multichar_Symbols
-Verb +Imp+P2-Sg +Imp+P2-PL +Imp+P3-Sg +Imp+P3-PL
64
+Pres
+PastIndef
+PastDef
+FutDef
+FutIndef
!+FutIndef+P1-Sg
+P1-Sg +P1-PL +P2-Sg +P2-PL +P3-Sg +P3-PL
+Neg
+Ques

LEXICON Root
Verb ;
Imp ;
Negat ;
NegatFutIndef ;
Pres ;
Past ;
PastPerf ;
FutDef ;
FutIndef ;
FutIndefPosit ;
PresSuffix ;
PastSuffix ;
FutSuffix ;
FutNegSuffix ;
X ;
V ;
Interog ;

!!!!!
! Lexical Entries
!!!!!

LEXICON Verb
bie V ; ! dance
Ayt V ; ! say
bar V ; ! go
cIrla V ; ! sing
CIk V ; ! come out
eSlA V ; ! work
eC V ; ! drink
kil V ; ! come
kien V ; ! dress
kit V ; ! go, leave
tap V ; ! find
tINla V ; ! listen
yarat V ; ! love
yaz V ; ! write
65
ukI V ; ! study, read
utIr V ; ! sit

!!!!!!
! Verb Morphemes
! Morphotactic Rules for Verbs: stem + Negation + Mood + Person/Number + Question
!!!!!!

LEXICON V
-Verb:0 Negat ;

LEXICON Negat
NegatPres ;
NegatOther ;
NegatFutIndef ;
FutIndefPosit ; ! The affirmative forms for FutIndef are different from the negated

LEXICON NegatOther
+Neg:7ma X ;
X ; ! Optional negation suffix

LEXICON NegatFutIndef
+Neg:7ma FutIndef ;!not an optional negation suffix for the future indef as the personal endings differ in affirmative and negative
forms

LEXICON NegatPres
+Neg:7m Pres ;
Pres ; ! Optional negation

LEXICON X
Imp ;
Past ;
PastPerf ;
FutDef ;

LEXICON Imp ! Imperative Mood Conjugates for 2nd and 3rd Person
+Imp+P2-Sg:0 # ;
+Imp+P2-PL:IgIz # ;
+Imp+P3-Sg:sIn # ;
+Imp+P3-PL:sInnar # ;

LEXICON Pres
+Pres:7Iy PresSuffix ;

LEXICON Past
+PastIndef:7gan FutSuffix ;

LEXICON PastPerf
66
+PastDef:7dI PastSuffix ;

LEXICON FutIndefPosit
+FutIndef:7Ir FutSuffix ; !after l,r,t -Ir , all other -ar, replace rule ; for affirmative form

LEXICON FutIndef
+FutIndef:7s PresSuffix ; ! only after negated form

LEXICON FutDef
+FutDef:7aCak FutSuffix ;

LEXICON PresSuffix
P1Sg ;
P2SgPres ;
P3Sg ;
PersPl ;

LEXICON PastSuffix
P1Sg;
P2Sg ;
P3Sg ;
PersPlPast ;

LEXICON FutSuffix
PersSg ;
PersPl ;

LEXICON FutNegSuffix
P1Sg ;
P2SgPres ;
P3Sg ;
PersPl ;

LEXICON PersSg
+P1-Sg:7mIn Interog ;
+P2-Sg:7sIN Interog ;
+P3-Sg:0 Interog ;

LEXICON PersPl
+P1-PL:7bIz Interog ;
+P2-PL:7sIz Interog ;
+P3-PL:7lar Interog ;

LEXICON PersPlPast
+P1-PL:7k Interog ;
+P2-PL:7gIz Interog ;
+P3-PL:7lar Interog ;
67
LEXICON P1Sg
+P1-Sg:7m Interog ;

LEXICON P2SgPres
+P2-Sg:7sIN Interog ;

LEXICON P2Sg
+P2-Sg:7N Interog ;

LEXICON P3Sg
+P3-Sg:0 Interog ;

LEXICON Interog
+Ques:7mI # ;
# ; !optional interrogatory suffix

#tatar-noun-rule.regex
#Noun Replace Rules; they are conditioned by suffix boundaries; ‘7’ indicates a morpheme boundary

[ l -> n || [mnN] 7 _ ]
.o.
[ d -> n || [mnN] 7 _ ]
.o.
[ d -> n || n I 7 _ a n ]
.o.
[ [..] -> n || I _ 7 d a .#. ]
.o.
[ g -> n || I 7 _ a ]
.o.
[ n -> [..] || n _ 7 Vowel .#. ]
.o.
[ g -> [..] || 7 I [mN] 7 _ ]
.o.
[ I -> [..] || I 7 n _ .#. ]
.o.
[ I -> [..] || [eIiaAo] 7 _ ]
.o.
[ s I -> t || I s 7 _ ]
.o.
[ [..] -> t || s 7 _ I ]
.o.
[ s I -> I || [ Cons | [ Vowel [uU]]] 7 _ ]
.o.
[ [uU] -> v || Vowel _ 7 I ]
.o.
[ k -> g, p -> b || Vowel _ 7 I ]
.o.
68
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _]
.o.
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _ , 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _ , 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ a -> A, o -> O, I -> e || 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ [ O A I o a ] -> [..] || Vowel _ ~.#. ]
.o.
[ y 7 [I|i] -> e || Vowel _ ]
.o.
[ g -> k, d -> t || VoicelessCons 7 _ ]
.o.
[ 7 -> [..] ] ;

# tatar-verb-rule.regex
#Verb replace rules; ‘7’ indicates a morpheme boundary

[ p -> b , k -> g || Vowel _ 7 ]

.o.
[ [..] -> y || a _ 7 a ]
.o.
[ l -> n || n 7 _ a r ]
.o.
[ s -> [..] || m a 7 _ 7 [m | b I z] ]
.o.
[ I -> a || Cons 7 _ r ] !FutIndefPosit
.o.
[ I -> [..] || I n _ .#. , m a 7 _ r ]
.o.
[ I -> [..] || [ e | I | a | A ] 7 _ g I z ] !Imperative Mood
.o.
[ Vowel -> [..] || Vowel _ r ]
.o.
[ I y -> a || Cons 7 _ ] !Present tense suffix
.o.
[ [Vowel|y] -> [..] || _ 7 I y ]
.o.
[ I y -> i || FrontVowel (y) (Cons)* 7 _ ] !Pres suffix
.o.
[ a -> [..] || m I y 7 _ ] !delete Present tense affirmative suffix
.o.
[ I -> i || FrontVowel (Cons)+ 7 m _ y ]
.o.
[ y -> [..] || FrontVowel (Cons)+ 7 m i _ ]
.o.
69
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _]
.o.
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _ , 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ a -> A, o -> O, I -> e || FrontVowel (y) (Cons)+ 7 (Cons) _ , 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ a -> A, o -> O, I -> e || 7 (Cons) FrontVowel (Cons)+ _]
.o.
[ d -> t , g -> k || VoicelessCons 7 _ ]
.o.
[ 7 -> [..] ];
70