6.4.1 Types of predicate
Formally, I distinguish four types of predicate: simple verbs, impersonals (in Pol-ish represented by forms ending with-noor-tonot inflected for person and num-ber, in Finnish according to the description of Hakulinen et al. 2004: §110) and the Finnish progressive construction with the third infinitive Inessive (see Section 4.5.1). Copulae are treated as a subtype of derivation.
6.4.2 Annotation of simple predicates and impersonals
An example of the Polish imperfective predicaterozgaduje si˛e‘become.3SG talk-ative’ looks like this:
<plverb>
<pllemma aspect="imperf" derivation="prefixed">
rozgadywa´c
</pllemma>
<plform>
<simple>
<num>sg</num>
<person>ter</person>
<tense>praet</tense>
</simple>
</plform>
<pfx>roz</pfx>
<plrefl>ref</plrefl>
</plverb>
Polish predicates are lemmatised to their infinitive forms stored in the element
<pllemma>which receives the attributeaspect, expressing the value ofPVA, and the attributederivationwhere the basic derivational types are given: cop-ula, simplex, prefixed, translative, and bare. The name of the prefix is given as a separate element. The Polish reflexive particle is annotated together with the verbal unit to which it belongs. The function of the reflexive particle is stored in the element<plrefl>.
Finnish verbal lexemes are annotated according to the particular derivational classes described in Section 4.2.1, that is: transitivisers (causative and attitude), detransitivisers (reflexives, reciprocals, decausatives, passive reflexives, transla-tives), and two temporal structure modifiers (frequentatives, momentaneous). Lex-emes not belonging to any of the above classes are classified as neutral.
Morphological information about verbs in text includes number, person and tense, according to the tag set used in the Wrocław CRF Tagger (Radziszewski 2013) for Polish and Fin-dep-par (Haverinen et al. 2014) for Finnish; impersonal forms include only the tense element.
6.4.3 Finnish progressive construction
The scheme contains a separate annotation of progressive construction (see Sec-tion 4.5.1) in the formAUX+INF3 in Finnish. The element<filemma>contains information about the infinitive lexeme, while the morphological information is stored in the<inf3>element.
<fiverb>
<filemma sem_group="causative">
todistaa
</filemma>
<fiform>
<inf3>
<aux>olla</aux>
<tense>Pres</tense>
<case>Ine</case>
<person>1</person>
<num>Plur</num>
</inf3>
</fiform>
</fiverb>
6.4.4 Nominal phrases in predicates
In the case of copulae, the structure of the nominal phrase of the predicate is stored in the<filemma> and<pllemma>elements. It preserves information about the type of phrase in line with the annotation scheme for non-predicate phrases (see Section 6.5.2) and the case (the Nominative is treated as the default case and is not annotated separately). The Finnish predicate phraseon hyvä‘be.3SGgood’
receives the representation:
<fiverb>
<filemma derivation="copula">NP</filemma>
<fiform>
<cop>
<num>Sing</num>
<person>3</person>
<tense>Pres</tense>
</cop>
</fiform>
</fiverb>
6.4.5 Free ad-verbals
Separate lexical units with free ad-verbals (see Section 4.2.1) present a big chal-lenge regarding the annotation of Finnish verbs. Therefore, the annotation scheme assumes the separate ad-verbal element<bounder>as in the example:
<bounder aggl="ill" syn="free">esiin</bounder>
embedding the information about ad-verbal morphology (attribute aggl) and syntactic position (attributesyn).
6.4.6 Light verbs and idioms
The next problem for annotation is caused by strong collocations with verbs, that is, idiomatic expressions and the light verbs constructions of the form: verb + noun phrase, such as the Polishwpu´sci´c w malinyliterally meaning ‘let somebody
into raspberries’, that is, ‘to fool’, or the Finnishsaada aikaan, meaning ‘to cause, create’.
As a solution, I annotated them within the verb element and tagged the collo-cate as a separate element<plfraz>or<fifraz>with a semantic label (see Section 6.5.4) of an argument or adjunct which it replaces, as in the expression dochodziły do głosu, literally ‘sound.PST.3PL’, idiomatically ‘become.PST.3PL
apparent’:
<plverb>
<pllemma aspect="imperf" derivation="prefixed">
dochodzic
</pllemma><base>chodzic</base>
<plform>
<simple>
<num>pl</num>
<person>ter</person>
<tense>praet</tense>
</simple>
</plform>
<pfx>do</pfx>
<plfraz role="direction">
<collocate>do glosu</collocate>
<pos>PrepNP</pos>
<case>do+gen</case>
<num>sg</num>
</plfraz>
</plverb>