2018-10-19

Language Without Metalanguage 3:
Discoveries Over Categories

Cases, and Aspects, and Moods, Oh, My!

Four steps, two rules. That is the essence of message parsing. But, as my title indicates, there are complications. So, sorry fam, but this is going to be fairly dry. In fact, unless you plan to use message parsing to teach languages or engage them at a higher (C1-minimum) level, just skip this one. This post merely closes some gaps that language experts will whine are open.

I'm not going to explain cases, aspects, and moods here. I'll just show some examples and work through them. If a sentence in some language seems impossible to message-parse, leave a comment, and I'll make a post or video on it.

What's the issue with cases?

The perfect seaman gift!
Message parsing breaks sentences into eight (or maybe nine) atomic sentences, not counting quantifiers. However, to some who see this list, it seems incomplete. Some sentences look like they require more arguments, and they look atomic. Here's one:
  • "I gave the seaman a gift."
If "the seaman" can't be removed, we introduce more atomic sentences. But, there's a kind of minimal functional completeness that learners should retain. Arbitrarily adding atomic sentences extends that minimum. That then makes message parsing less practical.

Most important, though, is that we would not see that these sentences mean the same thing:
  • "I gave the seaman a gift." ⇨ ADD and MOVE ⇨ "I gave a gift to the seaman."
That MOVE may not be available, though. Take German. Its dative structures appear with special determiners. If we can't ADD and MOVE them out of the way, we can still CUT them out:
  • "Ich gab dem Seemann ein Geschenk." ⇨ CUT ⇨ "Ich gab ein Geschenk."
We might also spot added arguments with morphemes. Take Russian and its case suffixes. We can CUT their extra dative arguments out, too:
  • "Я подарил матросу подарок." ⇨ CUT ⇨ "Я подарил подарок."
The main takeaway? We can't SPLIT or CUT parts off of atomic sentences. The conclusion? None of these sentences with extra arguments are atomic.

What about aspects and moods?

Aspects are less of an issue. In most cases, they require no parsing, at all.

Consider English speakers learning to use Chinese's "了". Shitty Chinese teachers say it's about tense, crappy Chinese teachers try to explain the perfective aspect, and decent Chinese teachers just show translated examples:
  • "我買車了。"
    → "I bought a car."
    → "I had bought a car."
  • “他來了。”
    → "He came."
    → "He has come."
    → "He is coming (now)."
Examples clearly beat parsing here. Also, trying to isolate aspects into atomic sentences requires those aspects to make it. That leads to an infinite regression. It's just one of those consequences of having to state linguistic metalanguage in an object language.

Moods, on the other hand, do need some explaining, since problems arise with irrealis moods. I'll stick to English here, since you'd have to know foreign irrealis structures to judge them:
  • "I would hate water if I were a sailor." ⇨ SPLIT ⇨ "I would hate water. I were a sailor."
You see the problem. Unfortunately, there's no ideal solution here. I patch message parses by marking them as non-sentences, and then transforming them into sentences with a substitution, like this:
  •  "I were a sailor." ⇨ "I could be a sailor."
A more technical way to handle it is to apply current syntax theory and to leave complement phrases as they are:
  • "I would hate water if I were a sailor." ⇨ SPLIT ⇨ "I would hate water. If I were a sailor."
Complement phrases aren't atomic, though, so you'll be back to a CUT step and a substitution, anyway.

Yeah, great, but I'm doing this for languages I don't know. Then what?

I'd recommend a language exchange partner or a bilingual friend. With them, you can verify that your parsed sentences are okay sentences. If you're learning by yourself, it's going to be much harder. It's not impossible, though, with a few pointers.

JFGI With Quotes and Asterisks

For those who don't know, JFGI means "just fucking Google it". What most Googlers don't know is that Google supports some regular expressions. So, for example, if I Google "just * Google it", I'll get these results. This matters for language learners because it can show which phrases are sentences.

Keep in mind, though, that the number of results is not very important. What's really important is the presence of results from native sources. That means that entries from Reverso, Linguee, TripAdvisor, and the like are not good indicators. Such sites archive or automatically generate translations. When the first page of results is full of such links, I jump to the sixth page. That's where I find actual uses, if they exist. If I'm message-parsing French, I'm looking for URL's like "sacrébleu.fr", a native French site for a French audience.

Also, keep in mind that rarer and longer sentences will probably have fewer results. I check pieces that are around four words in length. Too long, and there won't be any results. Too short, and the crap sites will clog the pages. Using a search engine wisely takes some practice.

Checking a Conjugator

If Googling doesn't work, I might check a conjugation guide. The Reverso Conjugator is useful for Romance languages. Learners should use them very sparingly, however. Wading through metalanguage is exactly what I want people to avoid. For most of you, if you find yourself analyzing a sentence this heavily, it's already a bad sign. Either the sentence is too difficult for you, or the sentence has a problem with it. Either way, I recommend the lazy way out.

When In Doubt, CUT It Out

CUT is the last step in message parsing for a reason: It's virtually fail-safe. You will lose parts of the narration. However, as a learner, your goal is to understand sentences. If stuff is in the way, just do as Uncle Joey says:

Good advice, Dave!
It's way simpler to Google phrases with words cut out of them. Also, you save yourself from wading into metalanguage, and that's the entire point. Learners' goals are not to know how languages work. They're to know what counts as language. Message parsing only comes in when something about a sentence's structure confuses a learner. Maybe the order doesn't make sense to them. Maybe they expect one structure, but encounter another. Whatever the case, it's not about analysis or categorization. It's about discovery. Message parsing is one route to that discovery. In my opinion, it's the route of least mental resistance. The amount of episteme (knowing-that) that most formal grammars impose gets in the way of the more important techne (knowing-how) of a language. So, it's best to make the parsing process painless, and, eventually, not to need it.

No comments:

Post a Comment