25 April 2006

Please Pass the Pragmatics

I've spent the last few weeks (lots of plane trips) reading Pragmatics by Levinson et al. Pragmatics is a difficult to define field (like many :P), but from a linguistic perspective, it is roughly "the study of the ability of language users to pair sentences with contexts in which they would be appropriate." (pg 24)

The book talks about five aspects of pragmatics:

  1. Deixis: methods of directly encoding context into language; "Meet me here a week from now with a stick about this big." (pg 55) Without knowing the context, we cannot know the meanings of me, now (and hence a week from now), this big. Diectic pronouns are pronouns that reference speaker- and hearer-known context.
  2. Implicature: Grice's theory says that a speaker should: be cooperative, be truthful (quality), be informative (quantity), be relevant, and be perspicuous (manner: avoid obscurity and ambiguity, be brief and orderly). (pgs 101-102) This theory is insufficient, but useful. Implicature also includes notions of quantification (perhaps, some, many, etc.) and metaphors.
  3. Presupposition: roughly describes that which is immediately inferrable but not the new information in an utterance. Eg., "Sue cried before she finished her thesis" presupposes that Sue finished her thesis, but this is not the new information (which is that she cried). (pg 187) There are both semantic and pragmatic presuppositions; the latter involve shared knowledge between speaker and hearer. (For fun, compare "Sue died before she finished her thesis.")
  4. Speech Acts: a speech act is a statement that does something rather than just one that says something (eg., "I declare war on Zanzibar." (pg 228)). The basic question of speech acts seems to be that of identifying them and understanding how they differ from normal statements.
  5. Conversational Structure: analysis of the sequential (and anti-sequential) nature of conversations, interruptions, etc.
I find implicature and presupposition fascinating, because this is essentially where hidden meaning comes out in text. (See, here I've presupposed that hidden meaning is fascinating!) I've really seen very few papers that look at computational aspects of these problems (speech acts and conversational structure may be exceptions). The textual entailment work, as far as I can tell, focuses on semantic implication rather than pragmatic implication, though of course this is related (but much easier, I would imagine!).

One immediate question is whether there is an appliation that demands pragmatic understanding (or, say, understanding of pragmatic implicature and presupposition). This, I'm not sure. I'm curious how divergent pragmatic issues are crosslingually. My hunch is "not much" which implies that this is not necessary for translation purposes (though Czech morphologically encodes for the new/old distinction). IR and IE also seem impervious to the pragmatic hammer. I can come up with artificial situations that suggest QA might benefit, but I feel these are too artificial (i.e., "Did Sue finish her thesis?"). Even summarization seems fairly robust to a lack of pragmatic understanding, although the "new/old" issue is important here. Perhaps what is old in the real discourse is not new for the reader of the summary. But if it just comes through as presupposition, it's unclear that anything is lost. I'm at something of a loss here, because it seems like human conversation has pragmatic influences so deeply embedded that it is surprising that I feel we can do without them for NLP problems.

12 comments:

Anonymous said...

In artificial intelligence, answering a "why?" question with a "because" answer may involve pragmatics, in the sense that the artificial mind has only its contextual background knowledge from which to dredge up a response which may or may not provide a logical answer such as we human beings are accustomed to. Like a child when asked why something happens, the AI Mind responds (pragmatically?) with the conjunction because and a statement of whatever occurs (is remembered) in association with the question. As artificial intelligence advances further, we may expect intelligent robots to respond with precise logic to human enquiries about causation.

Anonymous said...

How about the interpretation of indirect speech acts, such as "Can you pass me the salt?" (request for action instead of question for ability) or "Can you tell me the time?" (request for information instead of question for ability). These are questions that need to be interpreted beyond their literal meaning, and thus certainly a case of pragmatics essential to even a quite basic NLP/dialogue system.

hal said...

Nieske -- Are there so many cases like this that we're better off building a theory (which I believe would be quite complicated) and implementing it, rather than just writting down rules? In my very very limited experience, the "can you" expressions are quite similar in both English and Japanese; I wonder how much they diverge in other languages/cultures. ....or perhaps I'm being too pragmatic (ooh, word sense disambiguation!).

Nobuyuki Shimizu said...

Pragmatics shows up in situated language experiments. For example, a user gives an instruction to a robot.

"Take a right, then left. The second door on the right"

"After reaching the end of the hallway, turn right, the door is on the right."

Are you turning right into a hallway and then seeing a door, or are you turning right to face the door, so that you can enter it? To execute the actin, you need to know the difference.

M said...

Nobuyuki, I've been focusing on the pragmatics of following route instructions in exactly the sort of situated language experiments you're talking about. I'll be presenting this work at CogSci and AAAI this summer, but I'm not sure what the best NLP / CL forum for this is. Any suggestions?

Nobuyuki Shimizu said...

M, I was wondering about the same thing. Since it's in the dialogue and discourse category, ACL/HLT/SIG DIAL/SEM DIAL might be OK, but they are not exactly a best fit, like Hal said in this post.
http://nlpers.blogspot.com/2005/12/interesting-topics-in-nlp.html

Are there any other forums besides AAAI and CogSci? Robotics conferences?? I was thinking it has a better chance at general AI conferences than CL ones..

We are trying to use information extraction methods to understand route instructions. Drop me a line if interested. shimizu at cs. albany. edu.

hal said...

Nobuyuki -- I can easily believe that dialog systems or situated experiments run up against pragmatic issues reasonably frequently. Though the sorts of examples you've given seem to be just on the border of pragmatics -- essentially, they're pragmatical only in that their interpretation depends on context. But the advantage in this simple setting is that if the robot gets to the last action and sees a door, he can well assume one interpretation; if he does not, he can assume the other. (There is some difficulty if the robot's controller can make mistakes, which I suppose is often the case.) The advantage here is that the context is the world, while in many situations the context is the foregoing text. The latter seems harder because the interpretation problem is trickier. I'm curious if that is ever necessary.

Nobuyuki Shimizu said...

Hal,
I think the most natural case is whenever a request is implicit in a description.

It's hot. -> open the window.
I'm lost. -> tell me where it is.

If the request is complex, the description will be complex. The following is a dialogue that took place at a call center for a credit card company.

A man calls, and tell the operator that (1) he has lost a card, (2) he moved, (3) he doesn't know if his wife has called the call center about this or not.

The operator checked the status of the card, got his new address, cancelled the card being sent to his old address, and ordered a new card to be sent to his new address.

The caller didn't know what exactly his requests were, or how to express them, so instead he went about explaining his situation.

This is the kind of pragmatics you are thinking of, correct? In a call center application, there are many dialogue samples like these. Unfortunately, it's not easy for a company to release the data though. If we could do something statistically, it would be interesting..

Anonymous said...

very difficult and perhaps quite fallacious to compare natural langugage production to AI, not only because of the multitude of cognitive networking issues and multimodality, but also because spoken language comes under sexually selected pressure. amygdalic responses are preconscious (ie fast)and the upshot is that AI, even soft AI, has to be tweaked to allow for lack of gender identity

Anonymous said...

酒店經紀PRETTY GIRL 台北酒店經紀人 ,禮服店 酒店兼差PRETTY GIRL酒店公關 酒店小姐 彩色爆米花酒店兼職,酒店工作 彩色爆米花酒店經紀, 酒店上班,酒店工作 PRETTY GIRL酒店喝酒酒店上班 彩色爆米花台北酒店酒店小姐 PRETTY GIRL酒店上班酒店打工PRETTY GIRL酒店打工酒店經紀 彩色爆米花

Anonymous said...

自慰套,真愛密碼,
自慰套,自慰器,自慰套,情趣,充氣娃娃,
性感丁字褲,AV,按摩棒,電動按摩棒,情趣按摩棒,
角色扮演,角色扮演服,吊帶襪,丁字褲,飛機杯,

按摩棒,變頻跳蛋,跳蛋,無線跳蛋,G點,
潤滑液,SM,情趣內衣,內衣,性感內衣,情趣用品,情趣,

Luigi Speranza said...

Oddly, what one is passing is not really the salt, but the salt container! "salter" if you must. Surely it would be bad manners to touch the ingredient itself (the mineral) with your hands, no? Plus, little practical. JL Speranza. griceclub.blogspot.com. Grice, perhaps he had dwelt with Searle so much never paid proper attention to so-called indirect speech acts. I do like the pragmatics of 'please': "if it may please you" -- quite the opposite of what it 'says': it seems it's the receiver of the salt (container) who'll be eventually pleased, no? JL Speranza griceclub.blogspot