English Language & Usage: Multi-Layered Discourse Room

Cerberus

12:00 AM

Where else would they get that but from linguists?

RegDwight

@Kosmonaut Neither have I, but you never know.

Cerberus

The Eskimos probably have such words...

Kosmonaut

I have no clue; it would be a really bad idea to say that.

RegDwight

I mean, I assume our friend Cerberus here hasn't just overheard it during a soccer game.

Cerberus

@Kos: Well, why would that be so bad? It is basically the same as calling "until" a preposition—the same process is just applied much more rigorously or on a larger scale.

@Reg: I hate soccer! Well, playing it is OK; but the European Disease that is watching it, I deplore.

RegDwight

12:03 AM

Apr 1 at 0:50, by Cerberus

Cerberus

Haha I had no idea you were going to rake THAT back up.

For the record, that is not Cerberus!

RegDwight

I'm here to rake.

@Cerberus Yes, that is the guy who told you about prepositions.

Cerberus

His evil face does resemble a linguist's!

RegDwight

Evil? More like absent-minded.

Imagine Chomsky looking like that. Nobody would listen to him.

Cerberus

That's what he wants you to think! Look how the word "noun phrase" is on his lips!

Kosmonaut

12:04 AM

@Cerberus Maybe someone would make that theoretical claim, but then they would have to support it with evidence and predictive power.

But people who start treating multiword chunks as units nowadays are often not interested in the lexical categories at all

Cerberus

@Kos: I don't think it would be hard to support that: an adverbial phrase does function like an adverb. It is just a choice to call it that or not...?

Kosmonaut

@Cerberus You'd have to be able to do all the adverb tests and have it work that way... and there are "adverbial phrases" so maybe some things actually would fit in that category.

Cerberus

@Kos: Then how do you feel about calling "no matter" an adjunct, as in "no matter what I do, you will leave me"?

Kosmonaut

If you want to know how I feel... I hate current syntactic theory, which is why I study morphology and phonology

6

Cerberus

@Kos: But those tests are rather arbitrary: you just take what you already consider adverbs, find some criteria that they will all pass, and use that as a test; but the inclusion of adverbial phrases in the starting material determines what the criteria will be.

Kosmonaut

12:08 AM

There's no other way to do it.

Cerberus

Absolutely, agreed.

Kosmonaut

You have to define the term.

Cerberus

But if the adverbial phrases fail the definition or the test, they can simply be revised.

It is just a choice.

Kosmonaut

But then all of the old adverbs would fail the test.

Because they wouldn't do this extra thing.

Cerberus

?

I can just make a test "will they be able to function as adverbial constituents".

And scratch the old test.

Kosmonaut

12:11 AM

Maybe adverbial phrases do work fine already.

I mean, I know "in the evening" would be an adverbial phrase

for example

Cerberus

You mean the term "adverbial phrase" already works fine? I agree 100 %. But it is just a matter of choice whether we change the categories to include phrases in "adverb" or not.

Kosmonaut

But the words still have their own lexical categories

X-bar theory already has the notion of phrases

So a noun is actually called either an NP or a DP for Noun Phrase or Determiner Phrase

Cerberus

We could say "in" has no category for itself: it is just a prefix in some "adverbs".

Kosmonaut

No, it doesn't work like a prefix, it has to be a word.

Cerberus

Well, why?

Kosmonaut

12:14 AM

Because prefixes can't have anything get between them and the word they depend on.

So if "in town" has "in" as a prefix, you shouldn't be able to say "in the town" also.

That's just the definition of a prefix.

Cerberus

That would then just be consider an entirely different adverb.

RegDwight

Combinatorial explosion.

Kosmonaut

When "in the evening" is considered an adverbial phrase, it is not saying it is a word that actually is an adverb.

Cerberus

I know!

I am not saying you'd call it an adverb.

Nor that I'd call it that.

Kosmonaut

Besides, with "in", if you say "he sat in a chair", there is no adverbial phrase there.

Cerberus

12:17 AM

@Reg: You mean the number of possible combinations should be a criterion for considering something a lexically indivisible word, or several words? Seems like a pretty good criterion.

@Kos: Well, why isn't "in the chair" an adverbial phrase?

Kosmonaut

It is a prepositional phrase.

I gave the book to the man.

Cerberus

You mean if something is a complement it can't be adverbial?

Kosmonaut

Right

Although honestly, I really did give up on this level of deep analysis of syntax, because I hated it.

Cerberus

I think the dichotomy complement–adjunct isn't quite "hard" enough for it to rule out labels like "adverbial"...

RegDwight

@Cerberus Yeah, kinda. What I mean by "combinatorial explosion" is that the human brain obviously doesn't store "in the town" and "in town" as units. It stores "in", "the", and "town". Pinker mentions that, actually.

Cerberus

12:21 AM

@Kos: I feel your pain: the problem is, for me, that many of those categorizations only work up to a point: then the distinctions often becomes blurry/problematic. That is not to say they are useless; but we shouldn't take them for more than they really are.

Kosmonaut

Well, this is why Chomskyan syntax is in a tough spot these days.

But I honestly don't think there is a better way to categorize and analyze parts of speech than this.

Cerberus

@Reg: I think our brain stores units on several levels simultaneously: it stores both "in" and "in town", I think. That is why my intuition immediately tells me that "in town" is possible (it basically tells me that I have seen it before), but "at town" is not.

@Kos: Oh I think it is very valuable too.

RegDwight

@Cerberus Yeah well, I knew you would say that, because my wording wasn't the best.

Cerberus

Just as it is valuable to assume that we have a free will. But it isn't going to be as exact as mapping the endless and endless concordances that we actually use in our brains.

@Reg: Aww... but I do see your point.

RegDwight

A better example would be that it doesn't store "in the town", "in a town", "under the town", "under the tree" as units.

Kosmonaut

12:25 AM

@Cerberus The question is, how do people then create novel sentences?

And how do children start forming sentences when they have heard so few examples?

I've never said "above three otters", but I know it is proper English.

RegDwight

Seriously, Cerberus, get that introductory book by Pinker.

That would level the playing field.

Cerberus

@Reg: I think that it sort of does even that, to an extent... it is just that it is much more efficient to consider the building blocks of those phrases rather than all those endless phrases as wholes, because analysing the blocks will be good enough for predicting stuff. Exceptions need to be made, of course, and we call them idioms.

Kosmonaut

@Cerberus Some people do look at things that way, with some very compelling evidence. Frequency plays a huge role though. There is a tradeoff. So only really frequent phrases would get that kind of treatment

Cerberus

@Kos: Fair enough. We have the concordances in our brains, which we use passively to check phrases against and to take phrase out of to use; but we also analyse them to create new phrases.

Kosmonaut

But to a certain extent, the brain probably does do some of that.

Cerberus

12:28 AM

Agreed.

Kosmonaut

Also, when I say "how do people then create novel sentences?" I don't actually know the answer.

Chomskyan syntax is one attempt to answer that question.

RegDwight

Does anybody?

Kosmonaut

But it has all kinds of issues.

Bad issues.

Cerberus

@Kosmonaut Neither do I. I think people do use some kind of analysis too. And Chomsky helps. He just should be taken witha a grain of salt at least. Then again I don't know that much about this methods.

RegDwight

@Kosmonaut It's also too popular, so even where it could be quite on the mark, it gets misinterperted over and over again.

Cerberus

12:30 AM

Quite possible.

Kosmonaut

@Cerberus I think it was really useful for me to learn Chomskyan theory. It makes some really cool predictions that work in certain places. But then it really starts to crash and hit walls when you try to go all the way with it.

Most linguists aren't trying to do that anymore.

Cerberus

Yeah that is my impression as well.

Kosmonaut

Some are though, but I think their days are numbered, to be honest.

Still, the purely statistical models have also hit a wall.

Cerberus

If I had to guess, I'd say that much of the work our brain does to process/create language is done with brain functions that are also used for pattern recognition in other areas of life. The question is, how much?

@Kos: Isn't the wall simply the huge amount of context we use?

Every bit of knowledge in my brain I can use as context for my language.

Kosmonaut

@Cerberus Possibly, but how to integrate it?

I can say though, for speech recognition, the statistical models always outperform the tree models.

It's just that nobody can get past a certain level.

Cerberus

12:34 AM

Integrate it? I don't think the statistical method would need that: it just makes predictions with chance factor? And yes the Chomskyan method would seem to have no answer: it should just recognise its limits and happily continue to be useful wherever it is useful.

Kosmonaut

@Cerberus I mean, let's say you want to make something that can recognize speech using a statistical model.

Cerberus

Just as the free will is useful in prediction individual behaviour, though only up to a point; and much less so on the level of large groups of people and their interactions.

Kosmonaut

They can't recognize speech the way humans can.

Cerberus

Oh, yes. Perhaps huge amounts of data would spontaneously create the patterns we use for the same purpose in our brains?

Kosmonaut

It's not a matter of not enough statistical data.

We found that out about a decade ago — the statistical model tops out at like 95-98% accuracy.

You can't squeeze anything more out of it.

Cerberus

12:37 AM

What if we fed a computer the same amount of data a human brain receives in ten years?

Or perhaps a thousand times as much.

RegDwight

I think they feed computers way more than that.

Cerberus

Ehh I don't think so!

I don't know the bit rate of all the nerves that send input to the brain, but that would be huge.

RegDwight

They feed them entire corpora. Books none of us have ever read or will ever read.

Cerberus

Probably much more than is now technically possible to simulate...

Kosmonaut

I think we clearly use our understanding of what is going on and the meanings of words to predict what people are saying.

Cerberus

12:39 AM

Still, a thousand books surely contain much, much less data than what our eyes feed our brain in half an hour?

@Kos: Exactly.

Kosmonaut

@Cerberus It's just that the models can't do that job.

That stuff is not more of the same information, and you just need to catalog that too.

Cerberus

@Reg: I meant by "all data" literally all data, not just what we suck up in linguage data.

Kosmonaut

The model needs to incorporate some way to deal with that.

Robusto

You guys are still at it? Sheesh, and I thought I was the one without a life.

Cerberus

@Kos: Right, I have no idea what should be done with the data.

RegDwight

12:41 AM

@Cerberus Yes I get that now. But showing a computer a picture of a yellow truck won't teach it English, just as it won't teach me English.

Cerberus

@Rob: Think again! Sheesh.

Kosmonaut

@Cerberus That's where people are stuck.

Robusto

Reality is 80 million polygons per second. Fact.

Cerberus

@Reg: By my position would be that the picture of the yellow truck is a tiny brick of the construction of knowledge in your brain that allows you to pick up context.

Kosmonaut

So now in order to understand the meaning of what people are saying (let alone processing the visual cues), the computer needs a whole bunch more functionality. And does that functionality include some kind of tree-like parsing? Maybe, maybe not.

Cerberus

12:42 AM

@Kos: Yeah I have no idea how the problem could be solved practically, at all.

@Rob: polygons?

Robusto

In @RegDwight's case — the LEGO case — reality is about 80 polygons per second.

RegDwight

@Cerberus The thing is, the human brain actually works the other way round, for all we know. It ignores most of the information.

Robusto

@Cerberus — You don't know what a polygon is?

Kosmonaut

@RegDwight Yeah, it's still quite mysterious.

Cerberus

@Kos: Some kind of parsing of visual, etc. data would be necessary...

RegDwight

12:43 AM

There are so many things around you this very moment, and you can see them, but your brain ignores them and focuses on just a few pixels instead.

Kosmonaut

And at the moment, I don't think the neuroscientists can really talk to the theoretical linguists at all. There is a huge gap to close still.

Cerberus

@Rob: I know what a polygon is... but our sensory input is much more than polygons? Are you using it as a unit of data?

RegDwight

You don't understand my sentences by looking at your chair. You understand them by ignoring the chair.

Robusto

@RegDwight — That's kind of the definition of an online chat room. Sheesh.

RegDwight

Hey @Robusto, I pinged you a while back, can you check if that's a chiasmus or something else?

Cerberus

12:45 AM

@Reg: But if I say "is Rob sitting in a chair?", context will help me predict his answer, which is probably going to be "yes".

Robusto

@Cerberus — No. In virtual reality, it would take 80 million polygons per second to reproduce reality. About a gigabyte per second of bandwidth.

RegDwight

@Cerberus Dogs are cute.

Kosmonaut

@Cerberus Yeah, and with noise and lack of attention, you might not even hear some of the words

RegDwight

There, did anything in the world predict that answer?

I could say literally anything right now, completely out of any context imaginable, and you would understand it.

Robusto

@RegDwight — Chiasmus with a pun. You nailed it. +1.

RegDwight

12:47 AM

@Robusto Thank you kind sir. Because I honestly wasn't sure.

Cerberus

@Rob: Eh... I don't think I agree with that. What do polygons have to do with the sensory input my brain receives through its various nerves, including tactile, auditory, etc? And the polygons I perceive in vision are just an interpretation of my brain—not something inherent in the sensory data my visual nerves transmit to my brain.

@Reg: Not true! Nothing you say is ever without context.

RegDwight

@Cerberus Yes, but we don't always share the same context, you and I.

Cerberus

In fact, there is always a huge amount of context, namely what your brain and mine have as common patterns.

Robusto

@Cerberus — Rheingold (1991) quoted computer programmer Alvy Ray Smith as having said, "reality is 80 million polygons per second" (p. 168). This is reminiscent of filmmaker Jean-Luc Godard who once said, "Film is truth at 24 frames per second" (quoted in Pimentel & Teixeira, 1993, p. 23).

Kosmonaut

Still, you can say something out of context; it does happen

Robusto

12:48 AM

faculty.colostate-pueblo.edu/samuel.ebersole/336/eim/papers/…

Cerberus

@Rob: Okay if it is an expression, then your point is valid! Ding!

Robusto

Alvy Ray Smith

Alvy Ray Smith III (born 8 September 1943) is an American engineer and noted pioneer in computer graphics. Life and career In 1965, he received his bachelor's degree in electrical engineering from New Mexico State University. In 1970 he received a Ph.D. in Computer Science from Stanford University, with a dissertation on cellular automata. From 1969 to 1973 he was an associate professor of Electrical Engineering and Computer Science at New York University. While at Xerox PARC in 1974, he worked with Dick Shoup on SuperPaint, one of the very first computer paint programs. Smith's maj...

Cerberus

@Kos: I think we should take a very broad definition of context: in fact all the data in my brain are my context, of which the actual data exclusive to linguistic acts are only a small fraction.

Kosmonaut

Okay, yeah.

If you say that, then yes, arguably everything is probably in a context.

Cerberus

Dammit I wrote "data is"! How dare I!

Kosmonaut

12:51 AM

I always say "data is" unless I am writing a paper.

Robusto

@Cerberus — Fact: Everyone perceives things differently, filtered through the imperfect lens of their own perception, history, aspirations, emotional makeup (or in your case, real makeup). Haven't you ever seen Rashomon ferchrissakes? Sheesh!

Cerberus

@Kos: Yes, and I think a real machine that perfectly recognises speech by context would need that much context. The question is, how much of that would it need to be effective enough for our purposes?

RegDwight

LOL @ real makeup.

Cerberus

@Rob: can't disagree with that! And no what is Rashomon?

Robusto

@Kosmonaut — I say water are, because even the ocean is composed of little tiny droplets.

@Cerberus — What is Rashomon!!! Yer killin' me, doggie!

Kosmonaut

12:53 AM

@Robusto Then you're British. They like to do that kind of silly stuff.

Robusto

Rashomon (film)

is a 1950 Japanese crime mystery film directed by Akira Kurosawa, working in close collaboration with cinematographer Kazuo Miyagawa. It stars Toshirō Mifune, Masayuki Mori, Machiko Kyō and Takashi Shimura. The film is based on two stories by Ryūnosuke Akutagawa — ("Rashomon" provides the setting, while "In a Grove" provides the characters and plot). Rashomon can be said to have introduced Kurosawa and Japanese cinema to Western audiences, albeit to a small and discerning number of theatres, and is considered one of his masterpieces. The film won the Golden Lion at the Venice Film Fe...

Cerberus

@Rob: I'm sowwy, this dog is just not getting that many television channels down here...

Kosmonaut

@Cerberus The problem is (1) how to connect that info to the language info, and (2) how to process that information on the fly as people are speaking to you

Cerberus

@Kos: But I can't say "data is" because I am a Latin geek, d'oh!

Kosmonaut

@Cerberus Okay, fair enough :)

But I would call that an affectation more than a linguistic necessity :)

Robusto

12:54 AM

@Cerberus — So what do you say? Datums is?

Someone should hit the doggie with a Datumstempel.

Cerberus

@Kos: Yeah, a formidable problem. I just think focus should be on data that are not exclusively of a language kind.

I do say "agenda is", which pains me enough already!

Kosmonaut

@Cerberus I think there is focus on that... with people working on AI.

Cerberus

Good.

Kosmonaut

It's just really really tough to connect it to language at all!

Cerberus

I know!

RegDwight

12:56 AM

@Cerberus Then why do you keep trying?

Kosmonaut

I say "nonopus" for the plural of "octopus"

RegDwight

That's not plural. That's one and an eighth.

Robusto

So much of communication is context and expression, and a smaller part of that involves the actual words being said.

Cerberus

It is a bit of a hermeneutic circle: we use data to form our knowledge, and we use our knowledge to interpret the date, because we cannot convert the data to knowledge without interpreting them.

Robusto

Think of how many ways you can say "Dude" and have it mean something completely different each time.

Kosmonaut

12:57 AM

Dude!?

RegDwight

Duderino!

Cerberus

You mean enneapus? That is just a mollusc with nine arms!

Robusto

That's "El Duderino" to you. If you're, like, not into the whole brevity thing.

Dude.

RegDwight

But I am, hence no article.

Cerberus

@Rob: Very much agreed.

Robusto

12:58 AM

Duuuuuude.

Kosmonaut

Dewd

RegDwight

@Robusto That's the sound your phone makes when Peter Noone is on the other end.

Robusto

Dude?

Cerberus

Hey a question:

RegDwight

I like the direction our linguistics discussions always end up taking.

Robusto

12:59 AM

@RegDwight — I can't begin to fathom his existential pain when he picks up the phone. "Noone's home!" And he waits expectantly, every time, and they just ... hang ... up.

Transcript for

English Language & Usage: Multi-Layer…