5

Often languages have compounding phrases with the same Part Of Speech (POS) and it becomes a morphological analysis problem in Natural Language Processing (NLP).

The most notorious being infinite German noun compounds, e.g. Wissenschaftszeitvertragsgesetzausnahmeregelungsbeschluss which can be decomposed into Wissenschafts zeit vertrags gesetz ausnahmeregelung beschluss.

  • What other languages has such noun compounds?

  • Are there languages with the same phenomenon but with verbs? If so, what other languages has verb compounds? An example would be nice.

  • Are there open-source morpohological analyzers used for these languages to decompose the noun/verb compounds?

  • 1
    English has compound nouns, like "contract law exception control decision" although they're maybe not as common as in German. – dainichi Jul 14 '13 at 10:44
  • If you want to "decompose" this you should also decompose "ausnahme regelungs". – fdb Jan 2 '14 at 0:45
5
+50

You've already checked out wikipedia right? "Compound (linguistics)" Have you googled it on Google Scholar at all?

Noun compounds: This is common in Germanic languages. For instance, Norwegian "jernbanestasjonsmesterbolig" (jernbane stasjon mester bolig). Not a word you'll find in a dictionary but easily made when needed. The only thing that limits length is practicality. There are plenty of noun-noun compounds in English too but because the newly minted ones are in general spelled with spaces they don't stand out as much. For English you might want to start with Paula Downing (1977): "On the creation and use of English compound nouns", Language, issue 4.

For a smorgasboard of compounds, read up on Sanskrit.

Romance languages tend to compounds of the type verb+noun, like "tocadiscos" (toca discos) IIRC, "plays records" = record player.

Verb compounds: You've read "Compound verbs" on Wikipedia?

Analysis: The interesting, less homework-y question here is whether there are open-source morpohological analyzers out there now. Maybe one based on for instance Johannesen and Hauglin (1996) "An Automatic analysis of compounds"? You've alrady checked the NLTK I assume?

It's easy enough to make one yourself, for Germanic languages at least. You need a wordlist, potentially with declined forms, and if the word you have is not on the list, you try to match the words that are, as prefixes or suffixes, to the word you have. Go for longest match. (You'll also need some info on epenthetic sounds, like -e- and -s- in many Norwegian compounds. Throw away these, at least at first). The trick is that if there's more than two nouns you'll end up with potentially more than one tree-structure for the same word. Take "jernbanestasjonsmester", it can split at least two ways: "jernbane" + "stasjonsmester" and "jernbanestasjon" + "mester" (the -s- is epenthetic). Sometimes that doesn't change the meaning, sometimes it can change the meaning, sometimes it is used by comedians for comedic effect, setting up a joke that uses one split at the start and another later on in the same joke. And then there's the puns it makes possible...

| improve this answer | |
  • It's not as simple a problem as you've put it to build a German compound splitter, see github.com/alvations/DLTK/blob/master/tokenize/splicer.py ;) I thought it was simple too. – alvas Dec 28 '13 at 23:28
  • 2
    Of course it's not simple, language is involved :) Never expect 100% correct, be happy with 60% with easy techniques in NLP ;) For better than that you need to keep track of exceptions, use heuristics etc. It's like, real work or stuph. Seriously, if you get less than 60% doing the obvious stuff, try something radically different, as obvious won't do it. – kaleissin Dec 29 '13 at 0:17
2

1.Any polysynthetic language would have such a feature having within any of its word-sentences a 'verbal' particle. This is also one of the traits characteristical for French (with no distinction between parts of speech used for word formation, but with occasional differentiation between grammatical forms of the components), because this language may be considered as polysynthetic: prêt-à-porter (lit. ready-for-to wear, adj.+ preposition + noun), bien-dire (eloquence, lit. 'good-to speak'), bien-être (well-being, lit. 'good-to be'), or words like pince-sans-rire (a mocker with a poker face, lit. 'pinch-without-to laugh').

This principle of word-formation also includes combinations of noun+verb: un couvre-lit (bedspread, lit. '[you]cover-bed'), un tire-bouchon (corkscrew, lit. '[you] pull-cork'), un ouvre-boîte (can opener, lit. '[you] open-can') or those with plural imperatives (like cache-cache, formally and semantically (almost) equivalent to English hide-and-seek, ) or laissez-faire (negligence, lit. '[you (pl)] let-to be'), laissez-passer (a pass, lit. you (pl)] let-to pass').

This principle proliferates in structures of compound verbs with imperatives regarded as nouns, e.g. cessez-le-feu (a truce; lit. '[you (pl)] ceize-the-fire), décrochez-moi-ça (ragged clothes; a rag shop, lit. 'give-me-this [item of clothes hanging]'), or even le va-et-vient (the coming and going, lit. 'the [(s)he] comes ans [(s)he] goes').

2.Considering a verb as a form of action, we might regard some forms of Finnish and Estonian nouns as words-in-action, too, thus making such untypical (yet grammatical) structures as työhönmeno (to [active]work-going), poissaolo (absence. lit. 'away [stative]-being'), or maahanmuutto (immigration, lit. 'into land [active]-[causative] change') as specific examples of compound action-nouns.

3.In Russian, there is a peculiar feature which could be best described as '[ad]verbal predicates' with verbs with a form which looks much like a 2nd person sg. imperative: Это просто не-пришей-козе-баян. = This is strange and ridiculous (lit. 'This is a sheer don't-you-sew-an-acordeon-on-a-sheegoat').

4.In Chinese, where a distinction between parts of speech is arbitrary and depends on a syntactical structure, the word is regarded as a verb or as a noun depending on its place within a phrase. Thus, in 这是啥?(What is this) is 'is', or 'to be'; yet, although normally means 'suppling' or 'move' or 'at once', the combination of 就是 means 'exactly' or 'just like' rather than 'right-now-being' or 'being in a suppling motion'.

| improve this answer | |
  • any citation to prove that any polysynthetic language would have such a feature`? Even autronesian ones? – alvas Dec 29 '13 at 13:19
  • 2
    Any example of Austronesian polysynthetic language? Any grammar survey? – Manjusri Dec 30 '13 at 11:47
  • 1
    bien is an adverb, not an adjective. couvre is "it covers", not "you cover". – fdb Jan 1 '14 at 15:44
  • 'Bien' and 'couvre' are to be best understood within the general paradigm of adjective+verb (which is the case with other compounds) and 'sg. 2nd person imperative' VS 'pl. 2nd person imperative'. Otherwise there would be no grammatical system at all. – Manjusri Jan 1 '14 at 16:49
2

Nominal compounds are an inherited feature in Indo-European. Classical (i.e. post-Vedic) Sanskrit has lots of very long compounds rivalling anything in German.

| improve this answer | |
-1

Probably you might be looking for Agglutinative languages.

| improve this answer | |
  • No. agglutination means something different (the piling of some fixed set of affixes to a core word). – jk - Reinstate Monica Jun 30 '16 at 8:30
  • Those are fusional languages. – Suman Kumar Chalavadi Jul 4 '16 at 11:28
  • My mother tongue is Telugu, and I can see multiple words with their own stem/morpheme glue together with each other. That is agglutinate nature. – Suman Kumar Chalavadi Jul 4 '16 at 11:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.