* * Lecture notes by Edward Loper * * Course: Ling 590 (Pragmatics I) * Professor: Ellen Prince * Institution: University of Pennsylvania * Check the web page!! check on ling talks? email to manager@ling.upenn.edu \to ask to be added to the pennguists mailing list. [10/04/00 04:37 PM] Conventional terminology: semantic implication: imply, entails pragmatic implication: implicates > Horn Article reduce grice to two principles: Q-principle and R-principle. >> Q principle (quantity) minimize hearer effort (say what's sufficient, no less) "5 books" semantically implies \geq5, but pragmatically implies =5. "A or B" semantically implies OR, but pragmatically implies XOR. what about "he didn't read 5 books" (he read less than 5?) if he read 6 books\ldots >> R Principle (relation) minimize speaker effort (say what's necessary, no more) Explains things like "can you pass the salt?" Or "John was able to lift the car" meaning "John lifted the car" "John was able to lift the car" semantically implies he could.. "John was able to lift the car" pragmatically implies he did.. [10/18/00 05:41 PM] > Presentation on Clark 1979: Responding to Indirect Speech Acts How do you respond to a question like "do you know the time"? * Yes, I do. It's six. * It's six. >> Six Properties 1. Multiplicity of meaning: indirect speech acts have more than 1 meaning. 2. Logical priority of meanings: literal meaning is logically prior to indirect meaning. 3. Rationality: speaker and hearer must have mutual information 4. Conventionality: comvention of means & of form. Maximi: speak idionatically unless you have a reason not to. - can you pass me the salt? - could you pass the salt? - is it possible for you to pass me the salt? - are you able to pass the salt? - is it the case that you at present have the ability to pass the salt? 5. Politeness: use indirect speech acts to be polite 6. Purposefullness: listeners infer speech acts by recognizing goals, plans, and the roles speech acts play within plans. >> Responses to Indirect Speech Acts [10/25/00 04:34 PM] > Hobbes and Robinson Article: Why Ask? Addresses: what makes an answer appropriate? Speech is goal directed. In order to answer questions, you often need to be aware of their goals, in addition to the question itself. Answer is appropriate if: * it provides info to help questioner to achieve goals * if you don't have info helpful for goals, give 'most appropriate' response which will 'contribute most' towards the goals. > Allen Article: Recognizing Intentions From NL Utterances Model people as goal-directed agents, who maintain models of each others' goals etc. Define actions, goals, obsticles in terms of changing state of the world.. actions have preconditions & effects.. people construct plans: lists of actions, with parameters filled in for actions. Plans can be constructed with back-chaining and hierarchical decomposition. If you wnat to achieve P, and P might already be true, a goal becomes "know whether P is true." >> Plan Inference Agents attempt to infer other agents' goals & plans 1. infer goals, then simulate their planning process (how would I do that?) 2. observe their actions, and decide which plans it is consistant with. - If S observes A performing ACT, and ACT has effect E, then S may believe A is attempting to achieve E. - If S believes A wants to know whether P is true, S may believe that A wants to achieve P. >>> Obsticle Detection obstacles generally caused by lack of information required to execute a plan.. [11/01/00 05:42 PM] > Donnellan Article: Reference and Definite Descriptions >> What is Definite? - attributive: you believe that there is a reference that fits the description - referential: has a particular reference in mind. Consider "Smith's murderer is insane" - attributive: assert that whoever murdered Smith must be insane - referential: \exists X s.t. X murdered smith, X is insane.. If presuppositions are violated, referential definites still hold, attributive definites don't. Many definite descriptions are ambiguous between the two: I can talk about "the woman with glasses" but then later assert that I didn't mean -that- woman with glasses.. Distinctions are not always so clear.. "have your mother sign this" is attributive, but to the individual kids its sort of referential.. Proper names are mainly just referential. Using "whoever" forces attributive reading.. "whoever murdered John\ldots" Is >the king of france< degall? No, the president of france is degall\ldots hm. john is the king of france: "the king of france" isn't a new discource entity.. naming makes it easier to refer to things, but do names have any special place in language, or do they just pattern like other definite descriptions? > Prince Article: Toward a Taxonomy of Given - New Information Need a concept of old vs new info.. needs a good definition. Consider: # It's him that I like Presupposotion: I like someone; new info: it's him.. but him is 'old' as an entity, and 'new' as fitting into a particular slot.. Hearer's knowledge store (is it old/new for the hearer). It's new to the hearer in: # I saw a great movie this weekend.. Usually new info marked by being indefinite Also, we need to have discourse-old and discourse-new So we have a 2 dimentional matrix of new (at least): # Hearer-old Hearer-new # Discourse-old evoked xxxxxxxx # Discourse-new unused brand new Although noun phrases have systematic relations to these categories, they're not strict.. Consider: # I was on a bus the other day and we got into an accident # because _the_driver_ was drunk. The NP "the driver" isn't quite unused -- we know that it's the driver of the bus, etc.. Call these "inferables." Compare to: # I was on a bus the other day and we got into an accident # because _a_driver_ was drunk. This driver -can't- be the bus driver.. Inferables can be indefinite: # I was on a bus the other day and we got into an accident # because _a_wheel_ fell off. Containing Inferables: # Scholars such as E.C. and K.B. Evokes a set of scholars with a property (you get to infer). If you can't actually access the set, it's a containing inferable instead of an unused. [11/14/00 05:12 PM] > Structure and Ostension in the Interpretation of Discourse Deixsis [11/28/00 04:38 PM] blame/criticize: reversed assertion and presupposition\ldots projection problem -- how do presuppositions project up trees? > Karttunen Article: Presupposition and Linguistic Context Define satisfaction of presupposition to be a relation between context and sentences.. # satisfies_presup: context x sent \to bool For sub-clauses, the context can be expanded.. e.g.: # sat_presup(C, if A then B) = 1 iff: # sat_presup(C, A) # sat_presup(C\cup A, B) > Gazdar Article: A Solution to the Projection Problem