3.18. Hungarian Specifications

3.18.1.1. Derivation

Without handling derivation a satisfactory morphological analysis is not possible for Hungarian. The HUMOR system, a general purpose morphological analyzer, handles it and the results can be converted to Multext format since on syntactic level the morphological origin of stems are generally irrelevant. The resulting word class is defined by the rightmost derivational suffix. The suffix characters are literally attached to the word.

Here are the derivations that the analyser recognizes but instead of the origin we place only the resulting class to the output. (Suffix tags used in HUMOR are in upper case, actual suffixes are in lower case.)

Noun -> Adjective

Noun, A -> Noun

Noun -> Noun

Noun -> Verb

Verb -> Adjective

Verb -> Adverb

Numeral -> Adjective

Verb -> Noun

Adj -> Verb

Verb -> Verb

Examples:

If 'szemetelés' (littering, action of throwing away litter) is not in the dictionary we derive it from the verb 'litter' szemetel[V] + és[IF] (where IF=Verb2Noun).

Instead of giving the extra attribute to the verb expressing that it has a derivational suffix we simply give the result of the analysis+conversion: szemetelés[N].

In Hungarian some derivation may follow the inflectional suffix. For these derivations the suffix+derivation together forms a compound derivation. Then a new stem is generated from the stem + inflection + derivation segments and the resulting part of speech is determined by the derivation.

There are adverbs that may get case endings. Since case inflections derive adverbs from nouns these constructions can be handled as derivations. That means the stem is the stem + inflection combination and the part of speech is adverb.

3.18.1.2. Compounding

Compounding is handled in a very similar way to derivation. The rightmost word class is always the resulting one. If it contains some derivation as well then the result is the word class that the derivation determines.

Examples:

1. Hungarian is similar to German but we have a competence limit: we do not put together more then 2 nouns. These two words might be compounds as well but they must be lexicalized forms. So e.g.: rendőregyenruha (police uniform) is valid since it is put together from rendőr (police) and egyenruha (uniform) where rendőr=rend (order) + őr (guard) and egyenruha=egyen (uni) + ruha (clothes).
2. Of course the situation is more complicated since nouns can be derived from verbs, adjectives or other nouns, and these derivations can be parts of compounds: e.g. autóbuszmegálló (bus stop) where meg+áll+ó is derived from the verb áll (stand). Whereas megálló is lexicalized so dictionaries might contain it, the morphological rule is productive so compounding works well for constructions like tandíjbefizetés (paying the tution fee)= tan(tution)+ díj(fee)+ be(in) +fizet(pay) +és(ment).
3. Compounding, such as derivation, will not be mirrorred in the Multext format morphological analysis; a converter will produce an acceptable segmentation in the corpus. The word class (POS) for these compounds will be the word class of the output compound (e.g. tandíjbefizetés/N 'the action of paying in the tuition fee').