gstafleu

Just out of curiosity, do you have any examples of the two extremes?

Gerard Stafleu

Julian

I agree that my word usage has not been optimal in these posts so I will attempt to rectify that. I do not mean imply that there is any aesthetic difference between various grammars in objective sense.

First of all, one could argue that the existence of prescriptive grammar is a sign that some grammars are more acceptable than others. That doesn't necessarily mean better in any objective sense, I realize that, but there is a perception nonetheless.

Descriptive approaches to grammar cannot, by their very nature, impose any kind of judgement on the various data if they hope achieve some degree of accuracy.

When I talk about quantifiable efficiency and similar terms I am talking about a formal grammar, a phrase structure grammar of a type you could fit into a Chomsky hierarchy, for example, parse with LFG, SFG, DG, etc, etc... More later.
Easy in the sense that I have many ways of saying what I want to say, easy to make my words conform to my thoughts in a more precise manner.
I guess I am somewhat muddled. I though I clearly stated that there was no 'purpose' in the way that languages change other than what 'purpose' one might see in evolution, for example. I guess I better lay off that word because you seem intent on misunderstanding my usage of it. Anything that contributes to a viable system has a purpose, even if that purpose is only to keep that system viable. The use of the word purpose is not meant to imply intent or anything else that might indicate any kind of cognitive process.
My approach to grammar in most cases centers on formal grammar, as mentioned above. It is quite a simple matter to determine the number of variations, the number or rules, the manner of their execution, etc. It is quantifiable. How meaningful that might be is certainly open to debate. To a linguist, it's probably not very meaningful, understandably so, but for purposes of computational linguistics, it makes a big difference.

It also makes a difference to me personally when I speak a language, I can tell the difference between languages in terms of the expressive range possible, although I mostly feel this through the vocabulary. This is a personal observation that I would be wholly unable to provide evidence for.
As mentioned earlier, for descriptive linguistics it would be very difficult to have an opinion of better or worse and still avoid bias.
I wasn't actually doing that, I was merely clarifying my approach to this topic, but it does help me to know that you have a credentials in this field. I now know that when you are not understanding what I am saying, it is more likely to be my ability to explain myself than your ability to comprehend it.
I have more to say on this topic but I will leave off here for the moment and pick it up later since I have to go to a meeting. Yeah, they are actually making me work at work, bastards...

Sorry for any errors but I have no time to proof-read...

Julian

spin

Parenthetically, I doubt this statement. I think it's easier for your thoughts to conform to your words.


spin

Julian

That's not at all true. See for example,

Abney, Steven P. (1989) A computational model of human parsing. Journal of Psycholinguistic Research 18:129-144.

and correlate that with

Michael A. Covington. 2001. A fundamental algorithm for dependency parsing. Proceedings of the 39th Annual ACM Southeast Conference, pp. 95-102.

The above relates to similarities between human parsing and certain forms of machine parsing. I agree that expressive quality is not a useful term here. It does seems straightforward enough that more options means a range with finer granularity, possibly even greater range.

Julian

I am even less of a psychologist than I am a linguist, but I think I follow you here and you may be right.

Julian

Julian

Many languages have not been adequately described and no machine parser exists. In most cases this is because the investment is not worth it. Some languages that have complete free word order, or at least nearly so, are quite difficult to parse. Incidentally, there was no syntactical parser written for Koine Greek until very recently. Now there is one, constructed by yours truly. (It still needs some work for unusual constructions.)

Some languages are quite complex in their morphological capabilities (Turkish, for example) and others have word orders so free that it boggles the mind (I forget the names and I am not near my resources). If you like pain you can read through this dissertation: http://www.lashon.net/JMH/Publications/JMH-Thesis.pdf

English is the language where the most work has been done and parsers routinely score well into the 90+ percent in terms of accuracy. Strangely enough, most computational linguists who are not Americans (or British) also tend to work with English. A quick search on the web will show the availabilities of parsers and treebanks and the more you find, the better the language is understood, as a rule.

The Evil One

Heh, there is no situation on earth where I will read anything that long that begins with "I assume for concreteness the Minimalist program of Chomsky".

I have no time, but will try to respond to some of the other posts since last I was here shortly.

The Evil One

Right, this is going to be my last lengthy post. I copied Julian’s last response to Word and it came out as 3 pages. By the time I’d finished replying, which took about 2 hours all told, it was 7 pages. I just don’t have that much time, unfortunately, so this response is just going to have to stand for my position without any further lengthy elaborations. Julian – feel free to read my questions to you as rhetorical (I was intending not to write any but they just slipped out), I don’t demand answers when I know I won’t be able to respond at length.


Clearly so. But “acceptable” always means “acceptable TO SOMEONE”. The question is, who gets to do the accepting and decide what is or is not acceptable?

In this case, the existence of prescriptive grammar is a sign that some grammars are more acceptable than others to prescriptive grammarians. One is then faced with the question of why the PGs consider some forms better than others. The answer is always related to social prestige. So they consider “better” the forms that are associated with a prestige language (so, the form of English which is most similar to Latin is “better”) or with prestige dialects (so, the form of English associated with the dialect of South-East England is “better”) or with prestige contexts (so, the form of English used in writing is “better”).

But the reason that these forms are more highly valued is entirely social rather than linguistic, and largely a matter of historical accident. If the language of education in the Middle Ages had been Arabic rather than Latin, then a whole different set of English forms would be seen as superior by the PGs. If the dialect of English spoken by England’s rulers had been one of the ones which lack an adjective/manner adverb distinction, then you would right now be arguing that using the “-ly” morpheme was “unfortunate” and not as efficient for communication as leaving it off.


I’m not a Chomskyan (or any kind of formalist, in fact) and I’m not particularly familiar with the Chomsky hierarchy. However, you are surely aware that the position of Chomsky and his disciples is that all natural languages have basically the same grammar (that is, they all fall at exactly the same point in the Chomsky hierarchy). So you look in vain to Chomsky for any support for the idea that there can be a difference in “quantifiable efficiency” between any given pair of grammars.

But how do you quantify this "easiness"? How do you quantify the number of ways there are to say what you want to say? How do you quantify how easy it is to make your words conform to your thoughts? How do you quantify the precision of match between words and thoughts?
Again you claim that quantifying “the number of variations, the number or rules, the manner of their execution” is simple, without actually giving any hint as to how this is to be achieved. You don’t seem to understand how vast, and radical, a claim this is.



That’s the expressive range that you have access to, which of course differs from language to language. Unless you are one of the rare language prodigies, you probably have one or two native languages, one or two others you speak reasonably well, and maybe a handful of others that you are a bit familiar with. Your competence in these different languages varies so of course you will find yourself able to express more or less in different languages. The question is whether a similar difference can be found when compare like with like, i.e. the expressive range of language A for a native speaker of A, or the expressive range of B for a native speaker of B. Again, I don’t see where you would even begin to investigate it. What you see as easy I see as a wholly intractable problem.

(By the way, vocabulary is one of the areas where differences between the expressiveness of languages do exist and are detectable, e.g. if you’re speaking a language from a village in the highlands of Papua New Guinea then, chances are, your language won’t have all the words you’d need to write an instruction book on computer network maintenance or a scholarly article on nuclear physics, for instance. But we’re not talking about lexis. We’re talking about grammar.)


At this point, I should clear up a point which may or may not be a point of confusion. You admit that my point holds for “descriptive linguistics”, but what you don’t seem to acknowledge is that that’s the only real sort. Prescriptive linguistics – prescriptive grammar – is solely the domain of people who don’t really know anything about language except what they like and what they don’t like. These people often come up with complicated explanations of why what they like is objectively better in some way, these explanations often being very entertaining, but always being completely unfounded. This is what I see you doing with regard to “-ly”.



Well, the people I work with who do parsing (I’m familiar with some of Abney’s work, but not what you cite) are of the opinion that probabilistic parsing is the most effective. That’s what I was thinking of. Certainly there is no evidence that the human mind uses a hidden Markov model or anything like that.

Of course, knowing that you’re a formalist makes this point clearer. As I’m a non-formalist, I would say that human parsing can’t possibly be the same as machine parsing because machine parsing doesn’t use semantics, pragmatics, contextual information, and world-knowledge to aid the parse. But if you’re a formalist I imagine you’d argue that human parsing doesn’t either!

And again that claim, again without specifics. More options, but quantified how precisely? Finer granularity meaning what? Greater range of what? You have yet to answer any of my questions along these lines other than in airy generalities.


On Chinese plurals. I think we need a recap on what the point of the debate is here.

You claimed,

That is, you are claiming that English has a greater expressive range than Chinese because it has a plural inflection whereas Chinese has to indicate the plural with separate words. This was an explicit claim on your part.

To which I said,

and you replied,

Perhaps you can see how this looks, from my point of view, like a dodge. You have not addressed the question of why indicating the plural with a suffix rather than an extra word indicates “more options” or a “greater expressive range”. Instead you have introduced a whole load of red herrings. This does not impress me as to the strength of your argument. Still, let’s take those red herrings one at a time.


The unit in question is “the plural” so in English, it can be expressed in about seven main ways (/s/, /z/, /iz/, zero, ren, en, umlaut, plus some others in foreign loanwords) all of which mean exactly the same and are lexically or phonologically conditioned variants. I have no idea how many ways it can be expressed in Chinese but I’d guess it’s a lot more than seven, and I’d guess they’re not all synonyms (because you don’t get synonymy at the word level). So how exactly does that favour your argument?


And this diachronic notion is entirely irrelevant to what we were talking about.


No it’s not. Where do you get this from? You can’t make a word plural in English solely by the addition of some other word. The plural morpheme appears to be compulsory even if you’ve got another word present that indicates the plurality. So English can’t do everything that Chinese can. You are twisting the evidence to favour your conclusion.


But you didn’t answer my question about why it’s any better to have an affix than a separate word.


Have you never had to write -(s) at the end of a word to indicate that you’re referring to one or more than one? EG “poster(s)”. I know I’ve needed to do that. It’s a pretty clunky way of indicating unspecified number, isn’t it? For a start, how would you pronounce it? Almost makes you wish you spoke a language with some means of indicating unspecified number, eh?


Well, why didn’t you attempt to do this rather than rely on your own subjective impression, when arguing that possessing “-ly” is better than lacking “-ly”, or that having an inflectional plural is better than not having an inflectional plural?

Do you have any evidence that this so-called “objective method of quantification” actually produces the results you appear to claim it will?


I don’t get why you’re saying this. Didn’t you just get through saying that this measurement would be the only way to quantify the number of “options” a language has? And haven’t you been saying all along that the number of options available is what makes some forms of language better, in the sense of having greater “expressive range”? How exactly can a measurement, which, according to your own previous arguments, is the only way of objectively determining the “expressive range” of different languages, be “of dubious value”?

I think you are contradicting yourself.

I would actually agree that it’s of dubious value, but that’s because – as indicated above - I think formal grammar is, in its entirety, of dubious value when it comes to understanding how language actually works. So my agreement doesn’t really help your case.


Again you mention prescriptive grammar as if it actually has any value. It doesn’t. Prescriptive grammar consists of praising the language of groups of people you identify with and doing down the language of groups of people you don’t identify with. That’s all it consists of.

In the Bible shibboleth story, the Biblical writers recognised the role of language in delineating groups. But the soldiers in that story merely slew the poor souls who said “sibboleth”. They didn’t attempt to claim that their victims’ language was objectively lacking in expressive range or efficiency, relative to the “shibboleth” language of the victors. The latter approach is a modern conceit.


Please give a single example of a natural case in which the absence of a distinction between an adjective and a manner adverb has lead to “confusion”.

(Actually at this point I’d even settle for a contrived example, since we’ve not had any examples of any kind so far, but you do really need a natural example to prove the point.)

As I have indicated before, since the majority of languages lack this distinction, it should be extremely easy for you to find examples of this. You have, for a start, the entirety of everything ever written in German to scour for examples, since German (as I have mentioned) is one of the languages that lacks the adjective/manner adverb distinction.

As for the notion that “context and habit saves us” – well, if it’s not ambiguous in context, then no confusion can result, can it? That rather proves my point. Of course formalists love to pretend that context can be ignored, but the fact is that in the real world all language is interpreted in context.


Before leaving this issue, a fun fact: “-ly” (OE –lic) originally indicated an adjective, not a manner adverb – the cognate ending, “-lich”, still does indicate adjectives in German. It was more-or-less a matter of chance that led to it being reinterpreted as an adverb ending in English when the old adverb ending, “-e”, disappeared.


At a guess, because the number of people attempting to effect the change is almost always tiny compared with the number of speakers of the language, meaning that the inertia is very great. The imposed Latinisms, for instance, were never going to take hold because they were being imposed through an education system that the vast majority of English speakers weren’t exposed to until comparatively recently. That doesn’t mean it doesn’t work sometimes, e.g. the abolition of “generic he”. But then, “generic he” was another one of these imposed Latinisms so it wasn’t particularly rooted in the language anyway.


Spoken like a true formalist!

If you wanted to write about the rules according to which the plural is produced in, say, Farsi, how exactly would you discover those rules in the process of describing them? Obviously, you can’t. You’ve got to find out what they are. You’ve got to ask Farsi speakers, or analyse some Farsi text, to find out what they are before you can write about them.

Discovery and description are as separate in linguistics as they are in history or particle physics.

Only the most dyed-in-the-wool Chomskyans imagine that you can rely on introspection alone to tell you anything meaningful about language. Me, I take exactly the opposite view: although introspection is unavoidable, it can be dangerously misleading.

right - I'm done.

spin

Two issues, The Evil One, one more on topic than the other:

1) You wrote:

machine parsing because machine parsing doesn’t use semantics, pragmatics, contextual information, and world-knowledge to aid the parse

Things are slowly changing in this field at least in some sectors. How do you think parsers can distinguish between homonyms? Purely statistics?

2) On the "-ly" issue, I'll tentatively come up on Julian's side. Of course this suffix makes communication easier. It adds redundant information to the communication which means that the receiver of the information has to think a modicum less, or if they have missed a part of the communication they have an extra clue in reconstructing it: you know from the form of the word that it modifies a verb. Without the suffix you don't have this cue as to what it refers to and you must rely on other factors.

We lack a second person singular pronoun in English, which confuses the hell out of learners of the language, who are used to having the extra data. We are used to going without and compensate accordingly, but the adjustment that the learner has to go through shows that the extra data would have made the learning process easier.


spin

The Evil One

To reply briefly:
Often they can't. I don't know of any parser that can tell which of the "bank" homonyms is intended, for instance. Syntactic analysers make limited use of the linguistic context (e.g. using the presence or absence of "the" to distinguish nouns from verbs) but make no use of the non-linguistic context.

You'd think so, wouldn't you. If it's really the case, however, an example of confusion resulting from leaving off the "-ly" would not be hard to find. So far, I've not seen one.


Of course English would be easier to learn for French speakers if its grammar was more like French grammar. That proves nothing other than the fact that English and french are different languages. (Substitute whatever native language you like and the point remains the same). Your argument could be used to make exactly the opposite point, e.g.: