2

I was told here several times, that a part of speech is not universal, but specific for each language as much as the A,T,C and G's are in everyone's genome.

Nethertheless, occasionally the same terms are used for in different languages for practical reasons, e.g. Adjective in English and Japanese, even though they are similar but not equal in all respects.

But what if you compare the parts of speeches of thousands of languages? Is it still the case, that no two languages have identical parts of speeches or could it be, that there is a limit for specific ones?

To draw an analogy, each language possess a subset of all phonetic sounds possible across all languages as described in the IPA chart, but the number of all sounds possible would have a limit, even if millions of languages existed.

Is there something like an IPA of the parts of speeches or another approach to describe parts of speeches with a closed set of universal features?

  • 3
    As was pointed out in a comment to another recent question, the vowel space is continuous and not discrete, so in principle there are infinitely many vowel sounds. I think the case is similar for POS; there are no two languages where some POS acts exactly the same way, because there are simply too many parameters. – WavesWashSands Mar 21 '17 at 11:12
  • You might want to look at the nanosyntax / cartography program (Ramchand & Svenonius) which specializes in finely-splitting hairs to the molecular level. – user6726 Mar 21 '17 at 15:46
7

First of all, part-of-speech is not an observable. It is a latent category inferred from the utterances we can analyse.

As a latent category, it is dependent on our analysis. There are lots of parameters we can play with: Granularity of the analysis (some part of speech tagsets have more the 500 tags, simple models come with 8 or 10 parts-of-speech), input from favourite theories, inherited terminology and analysis from prior linguists, comparison to other languages or historical stages of the same language or a strictly internal analysis, remembering what we learned in school (the theories of yesterday, the new theories of the today's big shots are of course much better!). Different choices lead to different answers.

And yes, there are daring people trying to unify parts-of-speech over all languages of the world, most active are the people developing Universal Dependencies, see also this answer A list of parts of speech.

| improve this answer | |
  • Thank you, this was helpful, but I don't understand the meaning of the observable vs latent thing – Abdul Al Hazred Mar 21 '17 at 17:58
  • If you can observe that a word is a noun without looking at how it's used, that's observable. In tagged corpora, POS are observable -- but only to the extent that a particular tag set is satisfactory. In untagged corpora, one can often pick out POS (the is always an article, of a preposition, etc.) but not most of the time. If it can be determined by reasoning and further data, then it's latent (like a latent case of tuberculosis that can be discovered on an X-ray). Not all facts are obvious, and not all obvious conclusions are true. – jlawler Mar 22 '17 at 16:45
1

This is a theoretical question, so I would take my liberty to answer it in simple terms, too.

Nearly all sentences in all languages are about things, their (inter-)actions, and attributes (for both "things" and "actions"). This does not automatically guarantee that within an arbitrary language you will have discrete parts of speech serving the same purpose, but it would definitely suggest that there will be some classes of words that serve these three fundamental purposes.

Hence,

  • nouns and pronouns to denote things;
  • verbs to denote actions;
  • adjectives for attributes/properties of things;
  • adverbs for attributes of actions;

Further, each language has its own tools for delivering finer details of sense. These tools can be syntactical ones — separate parts of speech (like articles) in some languages, or morphological ones (like affixes).

| improve this answer | |
  • 2
    I think it's just a matter of terminology. Some languages, e.g. Manchu, can be analised as having no adjectives and no verbs, nouns can serve as attributes and predicates and they can have voice, tense and aspect forms. So it's just a terminological convention whether you call it a verb or a predicate noun. The Tagalog language is like that, too, "I read books" in Tagalog is expressed as "Real my reading of books" ­- no verb. – Yellow Sky Mar 21 '17 at 17:20

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.