Linguistics Study Guide Semantics

Notes- Semantics Structural Ambiguity Structural ambiguity = different meanings of a sentence depending on how you construct it • Ex: “I saw the spy with the binoculars” o N attachment: “with the binoculars” attaches to “spy”, modifying that noun to mean the spy has the binoculars o VP attachment: “with the binoculars” attaches to “saw the spy” and modifies that VP How semantics works • Assume simpler view with PS rules, meaning any well-formed expression can be a tree o tree just represents how syntax proves something well-formed: each PS rule is paired with a semantic rules • Sense vs reference o reference: what is picked out from the actual world o sense: meaning = the recipe for going from way the world might be to the individual (what you imagine, basically) ▪ if you were omniscient, you would know what it really means o NPs can pick out the same individual but the meaning/context is different ▪ ex: “the tallest man in the world right now” – we know what it means but not who it is referencing o VPs pick out sets of individuals ▪ • ex: dance is a subset of move because all dancers move Sentences o they tell us something about the world and are thus either true or false ▪ we can say that their referent (or extension) is a truth value ▪ set if truth values {true, false} or {1,0} o intension: a set of ways the world might be • Internal structure of NPs o “dog” = a set of individuals vs “a dog” = an individual 20 o “disobedient husky” = intersection of [[disobedient]] and [[husky]] ▪ • N → Adj N Adjectives o not all adjectives intersect with nouns ▪ ex: “the big flea” vs “the big elephant” shows that “big” means different things: they are subjective adjectives • size is always relative but need not be determined by the noun o relative/subjective adjectives ▪ uses comparison classes Picking it out • NPs- proper names, pronouns, and things pick out individuals • VPs pick out sets • Ss pick out truth values Truth values • Opinions have a truth value based on subjective parameters • Things are true or false with reference to a judge or a subject • Sentences pick out truth values • 0 = false, 1 = true • S1 → S2 and S3 o [[S1]] = 1 if and only if [[S2]] = 1 and [[S3]] = 1 o not circular—distinguish between object language vs metalanguage • S1 → S2 or S3 o [[S1]] = 0 if and only if [[S2]] = 0 and [[S3]] = 0 o In prose: S1 is false if both S2 and S3 are false, otherwise it is true Relative clauses • Relative clauses can modify N • N → N RC • Recursive rule because RC is a modifier/adjunct (same thing) 21 Determiners • “the” and “my” are examples of determiners • determiners take a set and picks out the unique or most salient member of the set o the relevant individual of a set o ex: [[husky]] is a set and [[the husky]] is an individual • NP → Det N, Det → the, some, every, many, few o complex determiner ex: at least three o quantificational determiners = determiners that specify an amount (such as few) Domain restriction • the final output of the linguistic system is usually underspecified and pragmatics come into play to give the final message o literal meaning: where the semantics leaves off o final message: interaction with the literal meaning, context, communication, etc • ex: my dog ate breakfast → means that the dog ate breakfast today • If the domain is widened, it has to also strengthen the sentence o adapted from Kadmon and Landman, 1993 Negative Polarity Items • words that only work in a negative sentence o negative sentence makes a restriction • Seemingly odd distribution • ex: any, ever, lift a finger, drink a drop, budge a single inch, give a hoot o they are idiomatic o they always occur with negatives • minimizers are a subclass of negative polarity items • “no” = empty intersection between phrases o ex: “no husky howls” = huskies don’t howl and howlers don’t include huskies • “every” = implies a subset o ex: “every husky howls” = huskies are a subset of the howler set 22 Upward entailing position • preserves the relation of semantic strength among a set of expressions (maintains truth value) • smaller to bigger • switching it to a superset allows it to preserve truth • ex: “some husky howled” → “some husky made noise” o VP position is upward entailing o doesn’t work the other way because it shrinks the possibilities • ex: “some husky howled” → “some dog howled” o N position is upward entailing o doesn’t work the other way because it shrinks the possibilities • the “→” should be able to translate to (“could lead us to reasonably believe that”) Downward entailing position • the opposite of upward-entailing: limits the range of possibilities by going from bigger to smaller • more common in negative sentences • a subset subs in and doesn’t guarantee the preservation of truth • Example o “every dog made noise” → “every husky made noise” o “Lee didn’t move” → “Lee didn’t dance” • “didn’t VP” - picks out the complement • the “→” should be able to translate to (“does not necessarily mean that”) • Downward entailing environments are environments where strength/info content is reversed o Ex: “some husky howled” is stronger than “some dog howled” or “some husky made noise” o no dog > no husky | every dog > every husky 23 Past tense and negation • Restricts the domain o within a relevant time interval, X did/didn’t happen o In the below example, it’s saying that “recently, I didn’t turn off the stove” where “recently” likely means that day • Ex: “Uh oh. I didn’t turn off the stove.” o Meaning 1: It’s not the case that there is a past time where I turned off the stove ▪ not correct semantics o Meaning 2: There is a specific relevant time where I didn’t turn off the stove ▪ translation: there is a past time t where it’s not the case that I turned off the stove • Ex: “I haven’t had breakfast” o Meaning 1: I have never in my life had breakfast o Meaning 2: I didn’t eat breakfast today ▪ Today is the interval it was restricted to “Ever” • Widens the interval – broadens the domain of times under consideration • Ex: “I haven’t eaten breakfast” means no breakfast today → “I haven’t ever eaten breakfast” means no breakfast at any point in life o widened the temporal window • If “ever” weakens the sentence or doesn’t add, it’s a bad sentence o Ex: I have eaten breakfast → I have ever eaten breakfast (bad sentence because it’s a negative polarity item added to a positive sentence) Semantics and syntactic categories • NPs -- pick out individuals (references) • VPs -- pick out sets • Adj -- pick out sets • Ss -- pick out truth values o S → NP VP, so [[S]] = 1 if and only if [[NP]] is a member of [[VP]] 24 Misc Semantics • double brackets = meaning/reference • metalanguage = language used to describe the semantics, which also happens to be in English • “Every” is bad in VPs and good in the NPs (?) • “Exactly 3” and similar things are not upward or downward entailing • “not” reverses the entailing patterns • If (a) is true, (b) is true → a > b • lexical semantics: words combining to create meaning • compositional semantics: phrases combing to create meaning 25