Wikipedia:Reference desk/Archives/Language/2018 June 15
| Language desk | ||
|---|---|---|
| < June 14 | << May | June | Jul >> | June 16 > |
| Welcome to the Wikipedia Language Reference Desk Archives |
|---|
| The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
June 15 edit
How many grammar rules are needed for speaking a language? edit
Every now and then I see lists of essential vocabulary for learners of (English as) a foreign language. They range from 1,000+ to 3,000 something. I wonder whether a list of essential grammar rules for nonnative learners would make sense too, and how long would this list be. It would for example include frequent structures like conditionals, relative clauses, but could exclude structures that can be circumvented like "He would have preferred to have been ..." or "Having been working for the government for a long time, my experience ..." --Hofhof (talk) 13:03, 15 June 2018 (UTC) Grammatical construction
- It seems that you're asking about constructions, more than about "rules" in the ordinary sense. Our Grammatical construction article is unfortunately quite brief... AnonMoos (talk) 13:37, 15 June 2018 (UTC)
- However, it links to Construction grammar, which has a lot more content. Loraof (talk) 15:19, 15 June 2018 (UTC)
- You could try looking at a descriptive grammar of English. Maybe you can even find one that numbers the constructions in English. Be aware, though, that authors may disagree on how many "rules" there are. Moreover, nesting and recursion can produce quite complex constructions. Linguists often study rules for constructing sentences, rather than specifying constructions. And, of course, speakers bend and break rules all the time. There is basically an unbounded realm of intelligible constructions. - Donald Albury 17:05, 15 June 2018 (UTC)
- Ain't it the truth! Bearing in mind that the most important rule is to be understood by your audience. ←Baseball Bugs What's up, Doc? carrots→ 18:32, 15 June 2018 (UTC)
- Tell that to James Joyce (the author of Finnegans Wake). -- Jack of Oz [pleasantries] 20:42, 15 June 2018 (UTC)
- In fact, his wife Nora Barnacle once asked him, "Why don't you write books people can read?". -- Jack of Oz [pleasantries] 08:06, 24 June 2018 (UTC)
- Tell that to James Joyce (the author of Finnegans Wake). -- Jack of Oz [pleasantries] 20:42, 15 June 2018 (UTC)
- Ain't it the truth! Bearing in mind that the most important rule is to be understood by your audience. ←Baseball Bugs What's up, Doc? carrots→ 18:32, 15 June 2018 (UTC)
- You could try looking at a descriptive grammar of English. Maybe you can even find one that numbers the constructions in English. Be aware, though, that authors may disagree on how many "rules" there are. Moreover, nesting and recursion can produce quite complex constructions. Linguists often study rules for constructing sentences, rather than specifying constructions. And, of course, speakers bend and break rules all the time. There is basically an unbounded realm of intelligible constructions. - Donald Albury 17:05, 15 June 2018 (UTC)
- Hofhof -- "Construction grammar" is one particular type of linguistic theory. If you want a summary of English constructions without a heavy dose of abstract theory, then you could look at A Communicative Grammar of English by Geoffrey Leech and Jan Svartvik, or The Syntactic Phenomena of English by James D. McCawley. If you're not afraid to go to an older source, there's always Otto Jespersen's A Modern English Grammar (in seven volumes)... AnonMoos (talk) 15:54, 16 June 2018 (UTC)
@Hofhof: It looks more and more like "rules" aren't that useful an approach to grammar. Machine translation is now done with deep neural nets suggesting that grammar is more a matter of correlation than "rules". This is an interesting popular-level article about Google Translate switching to a neural net approach. 173.228.123.166 (talk) 23:58, 23 June 2018 (UTC)
- 173.228.123.166 -- That's actually quite controversial. Neural networks are very good at scanning through vast quantities of text which is available in two languages, and discovering regularities and heuristics in how words and phrases are commonly translated between the two languages. Neural networks are very bad at understanding what they're "translating", and translators have known for a long time that a correct translation can't be guaranteed unless the translator understands the original text. AnonMoos (talk) 10:39, 24 June 2018 (UTC)
Not about English but auxiliary languages such as Esperanto uses 16 rules (at least in theory), while Interlingua may use a little more, but likely less than 50 or 100.--Lüboslóv Yęzýkin (talk) 11:17, 24 June 2018 (UTC)