Finnish Noun Inflection

The goal of this excercise is to build a lexicon that contains Finnish nouns inflected for number and case. To keep the problem reasonably small, we consider only three cases: Nominative, Genitive, and Partitive and two optional clitics. For the same reason, we only consider two types of nouns: monosyllabic and some bisyllabic stems. Even with these limitations, the problem is challenging because the shape of the endings depends on the shape of the stem and the stems are subject to several regular alternations such as Vowel Harmony and Consonant Gradation, and a number of other phenomena. This excercise assumes that the students is very familiar with xfst replace rules and the syntax of lexc source files.

The Facts

A Finnish noun begins with a stem. In all of the cases below, the stem is identical with the nominative singular. A plural marker, if any, immediately follows the stem. After the stem and the possible plural marker comes one of several possible case endings. We consider only three cases: Nominative, Partitive and Genitive. For the purpose of this excercise we, assume that the plural and case endings are the following:

Table 1: Plural, Genitive, and Partitive Markers
t	Nominative Plural marker
I	Plural marker for other cases, realized as i or j
n	Genitive Singular marker
Ten	Genitive Plural marker, realized as ten, den, or en
TA	Partitive Marker marker, realized as ta, tä, a, or ä.

There are numerous clitics that attach to the end of the word after the case ending. Here we only consider two of them.

Table 2: Clitics
kO	Question clitic, realized as ko or kö
hAn	Politeness clitic, realized as han or hän

The following table illustrates some of the possible combinations with the noun valo 'light'.

Table 3: Examples illustrating morphotactics with valo 'light'.
valo	Nominative Singular. No case ending
valot	Nominative Plural. The general plural marker I is not used in the nominative.
valoa	Partitive Singular. The partitive marker TA is realized here as a
valon	Genitive Singular
valoja	Partitive Plural. The plural marker I is realized as j here. The partitive marker TA is realized here as a.
valojen	Genitive Plural. The plural I is realized as j here. The Plural Genitive maker Ten is realized as en.
valoko	Nominative Singular with the question clitic.
valonhan	Genitive Singular with the politeness clitic.
valoakohan	Partitive Singular with the question and politeness clitics.
valojenkohan	Genitive Plural with the question and politeness clitics.

Note that the order of the two optional clitics is fixed. The form valohanko is a possible word in Finnish but it is interpretable only as a compound noun: valo 'light' + hanko 'pitchfork'.

Finnish is a language with vowel harmony. The realization of A and O in the first two tables depends on the previous "harmonizing vowel". Finnish has eight vowels: a, e, i, o, u, y, ä, and ö. The back vowels a, o, u and the front vowels ä, ö, y are harmonizing vowels, The two remaining ones, e and i, are neutral. For the purpose of this excercise, assume the following Vowel Harmony rule:

A -> a, O -> o || [a|o|u] ~$[ä|ö|y] _
.o.
A -> ä, O -> ö.

The realization of the T in the Genitive Plural and the Partitive marker depends on the syllable structure of the word. After a monosyllabic stem such as maa 'earth', the T is realized as t as in maata. After a bisyllabic stem such as valo 'light', the T disappears as in valoa. In cases where the T in Genitive Plural marker is realized as t, it is subject to the consonant gradation rules and surfaces as d in most cases.

The plural marker I is realized as j between two vowels, otherwise it is realized as i.

With this information you should be able to write the rules that correctly realize the six endings in all environments.

The remaining problem is that the stem of the noun also undergoes alternations. Consonant Gradation is one that we have already seen and solved. The other stem alternations, Vowel Rounding, Vowel Lowering, Vowel Dropping, and Vowel Shortening are illustrated in Table 4 below.

Table 4: Stem and suffix alternations
Nom Sg	Gloss	Nom Pl	Gen Sg	Part Sg	Gen Pl	Part Pl
puu	tree	puut	puun	puuta	puiden	puita
maa	earth	maat	maan	maata	maiden	maita
pää	head	päät	pään	päätä	päiden	päitä
syy	reason	syyt	syyn	syytä	syiden	syitä
pii	silicon	piit	piin	piitä	piiden	piitä
suo	swamp	suot	suon	suota	soiden	soita
työ	work	työt	työn	työtä	töiden	töitä
tie	road	tiet	tien	tietä	teiden	teitä

tikka	dart	tikat	tikan	tikkaa	tikkojen	tikkoja
pappi	priest	papit	papin	pappia	pappien	pappeja
hytti	cabin	hytit	hytin	hyttiä	hyttien	hyttejä
kukka	flower	kukat	kukan	kukkaa	kukkien	kukkia
tutti	pacifier	tutit	tutin	tuttia	tuttien	tutteja
kauppa	shop	kaupat	kaupan	kauppaa	kauppojen	kauppoja
kuoppa	hole	kuopat	kuopan	kuoppaa	kuoppien	kuoppia

jalka	foot	jalat	jalan	jalkaa	jalkojen	jalkoja
härkä	ox	härät	härän	härkää	härkien	härkiä
linko	sling	lingot	lingon	linkoa	linkojen	linkoja
kyky	talent	kyvyt	kyvyn	kykyä	kykyjen	kykyjä

sopu	harmony	sovut	sovun	sopua	sopujen	sopuja
kampa	comb	kammat	kamman	kampaa	kampojen	kampoja
piispa	bishop	piispat	piispan	piispaa	piispojen	piispoja

vahti	guard	vahdit	vahdin	vahtia	vahtien	vahteja
ilta	evening	illat	illan	iltaa	iltojen	iltoja
sota	war	sodat	sodan	sotaa	sotien	sotia
häntä	tail	hännät	hännän	häntää	häntien	häntiä

Vowel Rounding

Short a is rounded to an o in front of the plural marker I. Examples: tikkaa Partitive Singular, tikkoja Partitive Plural, kamman Genitive Singular, kampojen Genitive Plural, kauppaa Partitive Singular, kauppoja Partitive Plural. This does not happen if the vowel nucleus of the preceding syllable starts with a rounded vowel (o or u). See the rule for Vowel Dropping.

Vowel Lowering

Short i is lowered to e in front of the plural marker I. Examples: vahtia Partitive Singular, vahteja Partitive Plural, pappia Partitive Singular, pappeja Partive Plural. This does not happen in the Genitive Plural. See the rule for Vowel Dropping.

Vowel Dropping

A short a is deleted in front of the plural marker I if the nucleus of the preciding syllable consists of, or begins with, a rounded vowel (u or o). Note the different behavior of kuoppa where the a is dropped and kauppa where the a is rounded to o in the plural. Short ä is always deleted in front of the plural marker I. Examples: kukan Genitive Singular, kukkien Genitive Plural, sotaa Partitive Singular, sotia Partitive Plural, kuoppaa Partitive Singular, kuoppia Partitive Plural, härän Genitive Singular, härkien Genitive Plural. Stem-final i is dropped in the Genitive Plural. Examples: papin Genitive Singular, pappien Genitive Plural, vahdin Genitive Singular, vahtien Genitive Plural.

Vowel Shortening

In front of the plural marker I, the long vowels aa, ee, ii, oo, uu, yy, ää, öö, are shortened to a, e, i, o, u, y, ä, ö, respectively. The diphthongs uo, yö, and ie shorten to o, ö, and e, respectively. Examples: puuta Partitive Singular, puita Partitive Plural, tietä Partitive Singular, teitä Partitive Plural, työn Genitive Singular, töiden Genitive Plural.

The Task

Your task is to write a lexicon, a source file to lexc, that includes the 27 words, 4 suffixes and 2 clitics mentioned above and assembles them into morphotactically correct underlying Finnish forms. User the tags +Sg, +Pl, +Nom, +Gen, +Part for marking number and case on the lexical side, and +Q, +P for the question and politeness clitics. Compile using lexc or the command read lex in xfst. At this point your network should contain pairs such as

kukka+Sg+Part tutti+Pl+Gen jalka+Pl+Part härkä+Pl+Gen tie+Pl+Part
kukka TA tutti I Ten jalka I TA härkä I Ten tie I TA

Secondly, write replace rules for realizing the plural marker, the case endings, and the four stem-changing rules sketched above. Use the Vowel Harmony Rule shown above. Combine the suffix realization rules with the stem alternation rules sketched above. For consonant gradation, use the rule given in the Gradation script. Think about how to order the rules. It matters.

Finally, create an xfst script that reads in the lexicon, compiles the rules, and composes the lexicon with the rules leaving the result on the stack. The final result should contain pairs such as

kukka+Sg+Part tutti+Pl+Gen jalka+Pl+Part härkä+Pl+Gen tie+Pl+Part
kukka a tutt i en jalko j a härk i en te i tä

etc.

Verfify that the lower side of the lexicon contains the properly inflected surface forms by terminating the script with the command

print random-lower