Week 3a.  -roles, feature checking 3.5-3.6 CAS LX 522 Syntax I.

Slides:



Advertisements
Similar presentations
CAS LX 522 Syntax I Week 3b. Constituents.
Advertisements

Semantics (Representing Meaning)
NP Movement Passives, Raising: When NPs are not in their theta positions.
Syntax-Semantics Mapping Rajat Kumar Mohanty CFILT.
CAS LX 502 Semantics 10b. Presuppositions, take
Language and Cognition Colombo, June 2011 Day 2 Introduction to Linguistic Theory, Part 4.
L41 Lecture 2: Predicates and Quantifiers.. L42 Agenda Predicates and Quantifiers –Existential Quantifier  –Universal Quantifier 
Syntax Lecture 9: Verb Types 2.
Week 3b. Merge, feature checking CAS LX 522 Syntax I.
Grammatical Relations and Lexical Functional Grammar Grammar Formalisms Spring Term 2004.
CAS LX 502 8a. Formal semantics Truth and meaning The basis of formal semantics: knowing the meaning of a sentence is knowing under what conditions.
Week 3b. Constituents CAS LX 522 Syntax I.
Linguistic Theory Lecture 8 Meaning and Grammar. A brief history In classical and traditional grammar not much distinction was made between grammar and.
Term 2 Week 3 Semantics.
Episode 4b. UTAH CAS LX 522 Syntax I. Where we are We’ve just come up with an analysis of sentences with ditransitive verbs, such as Pat gave.
Introduction to Semantics To be able to reason about the meanings of utterances, we need to have ways of representing the meanings of utterances. A formal.
Episode 5a. TP CAS LX 522 Syntax I. Modals Pat might eat lunch. Pat might eat lunch. Modals: might, may, can, could, shall, should, will, would,
CAS LX 502 Semantics 1b. The Truth Ch. 1.
Constraining X-bar theory using the mental dictionary
Week 4. q Theory and the Big Picture
Installment 10b. Raising, etc CAS LX 522 Syntax I.
Week 9b. A-movement cont’d
Episode 8a. Passives and remaining issues CAS LX 522 Syntax I.
Week 8. Midterm debrief CAS LX 522 Syntax I. Midterm results Mean: 88 Mean: 88 Median: 93 Median: 93 A A- B+ B B-
Week 5b.  -Theory (with a little more binding theory) CAS LX 522 Syntax I.
Week 2a. Morphosyntactic features, part II. Ch. 2, 4.2- CAS LX 522 Syntax I.
CAS LX 522 Syntax I Week 10b. VP shells.
Installment 12a. Commentary, and the beginning of wh-movement ( ) CAS LX 522 Syntax I.
Episode 7b. Subjects, agreement, and case CAS LX 522 Syntax I.
1 CSC 594 Topics in AI – Applied Natural Language Processing Fall 2009/ Outline of English Syntax.
Meaning and Language Part 1.
Week 6a. Case and checking (with a little more  -Theory) CAS LX 522 Syntax I.
CAS LX 502 Semantics 2b. A formalism for meaning 2.5, 3.2, 3.6.
THE PARTS OF SYNTAX Don’t worry, it’s just a phrase ELL113 Week 4.
NLU: Frames Frame KR is a good way to represent common sense –can define stereotypical aspects of some domain we are interested in analyzing –sentences.
CAS LX 502 Semantics 3a. A formalism for meaning (cont ’ d) 3.2, 3.6.
Transitivity / Intransitivity Lecture 7. (IN)TRANSITIVITY is a category of the VERB Verbs which require an OBJECT are called TRANSITIVE verbs. My son.
Introduction to English Syntax Level 1 Course Ron Kuzar Department of English Language and Literature University of Haifa Chapter 2 Sentences: From Lexicon.
Syntax Lecture 8: Verb Types 1. Introduction We have seen: – The subject starts off close to the verb, but moves to specifier of IP – The verb starts.
The Lexicon Constraining X-bar theory using the mental dictionary.
CAS LX 502 8b. Formal semantics A fragment of English.
October 15, 2007 Non-finite clauses and control : Grammars and Lexicons Lori Levin.
Linguistic Essentials
Lecture 17 Ling 442. Exercises 1.What is the difference between (a) and (b) regarding the thematic roles of the subject DPs. (a)Bill ran. (b) The tree.
Albert Gatt LIN3021 Formal Semantics Lecture 4. In this lecture Compositionality in Natural Langauge revisited: The role of types The typed lambda calculus.
Week 3b. Merge, feature checking CAS LX 522 Syntax I.
Rules, Movement, Ambiguity
1 LIN 1310B Introduction to Linguistics Prof: Nikolay Slavkov TA: Qinghua Tang CLASS 16, March 6, 2007.
Making it stick together…
Sight Words.
SYNTAX.
3 Phonology: Speech Sounds as a System No language has all the speech sounds possible in human languages; each language contains a selection of the possible.
◦ Process of describing the structure of phrases and sentences Chapter 8 - Phrases and sentences: grammar1.
Lecture 1: Trace Theory.  We have seen that things move :  Arguments move out of the VP into subject position  Wh-phrases move out of IP into CP 
Week 3a.  -roles, feature checking CAS LX 522 Syntax I.
Lec. 10.  In this section we explain which constituents of a sentence are minimally required, and why. We first provide an informal discussion and then.
Week 3. Clauses and Trees English Syntax. Trees and constituency A sentence has a hierarchical structure Constituents can have constituents of their own.
Week 2. Constituents Sentences can be analyzed into subparts which we referred to as constituents English Syntax.
Week 3b. Merge, feature checking
English Syntax Week 12. NP movement Text 9.2 & 9.3.
Constraining X-bar theory using the mental dictionary
Syntax Lecture 9: Verb Types 1.
ARTIFICIAL INTELLIGENCE
Part I: Basics and Constituency
ENG 3306 Raising and Control I.
Linguistic Essentials
Valence, Transitivity, Voice
Structure of a Lexicon Debasri Chakrabarti 13-May-19.
Syntax Lecture 12: Extended VP.
Presentation transcript:

Week 3a.  -roles, feature checking CAS LX 522 Syntax I

Previously, in LX522… We left off last time exploring the idea that sentences are built from syntactic objects by using the operation Merge, taking two syntactic objects and forming a new one from them. We left off last time exploring the idea that sentences are built from syntactic objects by using the operation Merge, taking two syntactic objects and forming a new one from them. Big picture: What we’re trying to model is a system that can construct all—and only—those strings of words that correspond to sentences of a language (e.g., English). If we succeed, this system is (at least isomorphic to) what we know when we know the language. Big picture: What we’re trying to model is a system that can construct all—and only—those strings of words that correspond to sentences of a language (e.g., English). If we succeed, this system is (at least isomorphic to) what we know when we know the language. BC D E

Previously, in LX522… So far, we have: So far, we have: The lexicon (containing words, bundles of features) The lexicon (containing words, bundles of features) Merge (forms a syntactic object from two others). Merge (forms a syntactic object from two others). Merging two objects yields a new object that has the properties of one of the two objects. Merging two objects yields a new object that has the properties of one of the two objects. Merging eat and lunch yields a object that has the same kinds of properties as eat. Merging eat and lunch yields a object that has the same kinds of properties as eat. The object whose features determine the features of the new object ( project ) is the head of the new object—the most important component. The object whose features determine the features of the new object ( project ) is the head of the new object—the most important component. The question now is: how does Merge know which one is the head? The question now is: how does Merge know which one is the head? BC D E

Who’s in charge here? The idea we are going to pursue here is that one of the two objects needs Merge to happen—and the needy one is the head. The idea we are going to pursue here is that one of the two objects needs Merge to happen—and the needy one is the head. What does it mean to “need to Merge”? What does it mean to “need to Merge”? Consider hit. This can’t really stand on its own. It doesn’t mean anything (its truth can’t be evaluated) without providing a hitter and a hittee. Consider hit. This can’t really stand on its own. It doesn’t mean anything (its truth can’t be evaluated) without providing a hitter and a hittee. BC D E

Predicates, arguments, and propositions Conventional wisdom has it that a sentence needs a subject and a predicate. Conventional wisdom has it that a sentence needs a subject and a predicate. The idea is that the sentence expresses that the property signified by the predicate holds of the subject. The idea is that the sentence expresses that the property signified by the predicate holds of the subject. Pat danced. Pat danced. Danced is the predicate, it’s a property that Pat, the subject, has (if the sentence is true). Something that can be true or false, a “complete thought”, is a proposition. Danced is the predicate, it’s a property that Pat, the subject, has (if the sentence is true). Something that can be true or false, a “complete thought”, is a proposition.

Verbs and arguments Some are basically complete as they stand. Some are basically complete as they stand. Rain: It rained. Rain: It rained. Some have only a subject, they can’t have an object—the intransitive verbs (1-place predicates). Some have only a subject, they can’t have an object—the intransitive verbs (1-place predicates). Sleep: Bill slept; *Bill slept the book. Sleep: Bill slept; *Bill slept the book. Some also need an object—the transitive verbs (2-place predicates). Some also need an object—the transitive verbs (2-place predicates). Hit: *Bill hit; Bill hit the pillow. Hit: *Bill hit; Bill hit the pillow. Some need two objects— ditransitive verbs (3- place predicates). Some need two objects— ditransitive verbs (3- place predicates). Put: *Bill put; *Bill put the book; Bill put the book on the table. Put: *Bill put; *Bill put the book; Bill put the book on the table.

Verbs and arguments The “participants” in an event denoted by the verb are the arguments of that verb. The “participants” in an event denoted by the verb are the arguments of that verb. Some verbs require one argument, some require two arguments, some require three arguments, some require none. Some verbs require one argument, some require two arguments, some require three arguments, some require none. Intuitively, the number of arguments is the number of things that a verb needs in order to make a proposition (something that can be either true or false). Intuitively, the number of arguments is the number of things that a verb needs in order to make a proposition (something that can be either true or false).

Predicates We will consider verbs to be predicates which define properties of and/or relations between the arguments. We will consider verbs to be predicates which define properties of and/or relations between the arguments. Bill hit the ball Bill hit the ball There was a hitting, Bill did the hitting, the ball was affected by the hitting. There was a hitting, Bill did the hitting, the ball was affected by the hitting. Different arguments have different roles in the event. (e.g., The hitter, the hittee) Different arguments have different roles in the event. (e.g., The hitter, the hittee)

Thematic relations It has come to be standard practice to think of the selectional restrictions in terms of the thematic relation that the argument has to the verb—the role it plays in the event. It has come to be standard practice to think of the selectional restrictions in terms of the thematic relation that the argument has to the verb—the role it plays in the event. One thematic relation is agent of an action, like Bill in: One thematic relation is agent of an action, like Bill in: Bill kicked the ball. Bill kicked the ball.

Common thematic relations Agent : initiator or doer in the event Agent : initiator or doer in the event Theme : affected by the event, or undergoes the action Theme : affected by the event, or undergoes the action Bill kicked the ball. Bill kicked the ball. Experiencer : feel or perceive the event Experiencer : feel or perceive the event Bill likes pizza. Bill likes pizza. Proposition : a statement, can be true/false. Proposition : a statement, can be true/false. Bill said that he likes pizza. Bill said that he likes pizza.

Thematic relations Goal : Goal : Bill ran to Copley Square. Bill ran to Copley Square. Bill gave the book to Mary. (Recipient) Bill gave the book to Mary. (Recipient) Source : Source : Bill took a pencil from the pile. Bill took a pencil from the pile. Instrument : Instrument : Bill ate the burrito with a plastic spork. Bill ate the burrito with a plastic spork. Benefactive : Benefactive : Bill cooked dinner for Mary. Bill cooked dinner for Mary. Location : Location : Bill sits under the tree on Wednesdays. Bill sits under the tree on Wednesdays.

Thematic relations Armed with these terms, we can describe the semantic connection between the verb and its arguments. Armed with these terms, we can describe the semantic connection between the verb and its arguments. Ray gave a grape to Bill. Ray gave a grape to Bill. Ray: Agent, Source, … Ray: Agent, Source, … A grape: Theme A grape: Theme Bill: Goal, Recipient, … Bill: Goal, Recipient, …

Intransitives: Unergatives and unaccusatives For intransitive verbs (1-place predicates), there are two primary classes: For intransitive verbs (1-place predicates), there are two primary classes: Unergatives: Agent assigned to argument. Unergatives: Agent assigned to argument. Pat danced. Pat yodelled. Pat danced. Pat yodelled. Unaccusatives: Theme assigned to argument. Unaccusatives: Theme assigned to argument. Pat tripped. The boat sank. Pat tripped. The boat sank. Basically what you’d expect considering the normal transitive (2-place) verbs that have an Agent and a Theme. Basically what you’d expect considering the normal transitive (2-place) verbs that have an Agent and a Theme. The naming of these classes is not my fault. The naming of these classes is not my fault.

 -roles An argument can participate in several thematic relations with the verb (e.g., Agent, Goal). An argument can participate in several thematic relations with the verb (e.g., Agent, Goal). In the syntax, we assign a special connection to the verb called a “  -role ”, which is a collection of thematic relations. In the syntax, we assign a special connection to the verb called a “  -role ”, which is a collection of thematic relations. For the purposes of syntax, the  -role (the collection of relations) is much more central than the actual relations in the collection. For the purposes of syntax, the  -role (the collection of relations) is much more central than the actual relations in the collection.  -role Agent Source

 -roles We will often need to make reference to a particular  -role, and we will often do this by referring to the most prominent relation in the collection. We will often need to make reference to a particular  -role, and we will often do this by referring to the most prominent relation in the collection. For example, in Bill hit the ball, we say that Bill has the “Agent  -role”, meaning it has a  -role containing the Agent relation, perhaps among others. For example, in Bill hit the ball, we say that Bill has the “Agent  -role”, meaning it has a  -role containing the Agent relation, perhaps among others.

The Unique  Generalization Each  -role must be assigned to a constituent, but a constituent cannot be assigned more than one  -role. Each  -role must be assigned to a constituent, but a constituent cannot be assigned more than one  -role. (a.k.a. “the  -criterion”) (a.k.a. “the  -criterion”) Verbs have a certain number of  -roles to assign (e.g., say has two), and each of those must be assigned to a distinct argument. Verbs have a certain number of  -roles to assign (e.g., say has two), and each of those must be assigned to a distinct argument.

C-selection Verbs are recorded in the lexicon with the  - roles they assign as part of their meaning. Verbs are recorded in the lexicon with the  - roles they assign as part of their meaning. But, (some) verbs can assign the same type of  -role to different categories of argument: But, (some) verbs can assign the same type of  -role to different categories of argument: Pat felt a tremor. Pat felt uncomfortable. Pat felt that Chris had not performed well. Pat felt a tremor. Pat felt uncomfortable. Pat felt that Chris had not performed well. Pat is the Experiencer; a tremor (noun), uncomfortable (adjective), or that…well (sentence) is the Theme/Source. So  -role does not determine syntactic category. And syntactic category certainly does not determine  -role. Pat is the Experiencer; a tremor (noun), uncomfortable (adjective), or that…well (sentence) is the Theme/Source. So  -role does not determine syntactic category. And syntactic category certainly does not determine  -role. Pat kicked a pail. *Pat kicked unhappy. *Pat kicked that the earth is round. Pat kicked a pail. *Pat kicked unhappy. *Pat kicked that the earth is round. So verbs also need to be recorded with information about the syntactic categor{y/ies} they combine with. So verbs also need to be recorded with information about the syntactic categor{y/ies} they combine with.

C-selection (“Subcategorization”) Kick needs a nominal object. Kick needs a nominal object. Pat kicked the pail. Pat kicked the pail. Kick has a [V] category feature, but also needs to have an [N] category feature in some form to specify that it needs a nominal object. Kick has a [V] category feature, but also needs to have an [N] category feature in some form to specify that it needs a nominal object. BUT—We don’t want to risk interpreting kick as a noun, though. So, the [V] and [N] features must have a different status. BUT—We don’t want to risk interpreting kick as a noun, though. So, the [V] and [N] features must have a different status. On kick, the [V] feature is interpretable. The [N] feature is just for use in assembling the structure, it is not interpreted—hence uninterpretable. On kick, the [V] feature is interpretable. The [N] feature is just for use in assembling the structure, it is not interpreted—hence uninterpretable. The uninterpretable feature is an instruction for Merge. The uninterpretable feature is an instruction for Merge. The interpretable feature plays a role in determining the meaning of the word. The interpretable feature plays a role in determining the meaning of the word.

C-selection Not all transitive verbs (that take just one object) can take the same kind of object. Not all transitive verbs (that take just one object) can take the same kind of object. Sue knows [ DP the answer ] Sue knows [ DP the answer ] Sue knows [ CP that Bill left early ] Sue knows [ CP that Bill left early ] Sue hit [ DP the ball ] Sue hit [ DP the ball ] *Sue hit [ CP that Bill left early] *Sue hit [ CP that Bill left early] So know can take either a DP or a CP as its object argument; hit can only take a DP as its object argument. So know can take either a DP or a CP as its object argument; hit can only take a DP as its object argument.

S-selection Verbs also exert semantic control of the kinds of arguments they allow. Verbs also exert semantic control of the kinds of arguments they allow. For example, many verbs can only have a volitional (agentive) subject: For example, many verbs can only have a volitional (agentive) subject: Bill likes pizza. Bill kicked the stone. Bill likes pizza. Bill kicked the stone. #Pizza likes anchovies. #The stone kicked Bill. #Pizza likes anchovies. #The stone kicked Bill. We’ll assume that this is not encoded in the syntactic features, but if you mess up with respect to s-selection, the interpretation is anomalous. We’ll assume that this is not encoded in the syntactic features, but if you mess up with respect to s-selection, the interpretation is anomalous.

Feature checking For our model, we will say that if a syntactic object has an uninterpretable feature, it must Merge with a syntactic object that has a matching feature — and once it’s done, the requirement is met. The uninterpretable feature is checked. For our model, we will say that if a syntactic object has an uninterpretable feature, it must Merge with a syntactic object that has a matching feature — and once it’s done, the requirement is met. The uninterpretable feature is checked. Specifically: Specifically: Full Interpretation: The structure to which the semantic interface rules apply contains no uninterpretable features. Full Interpretation: The structure to which the semantic interface rules apply contains no uninterpretable features. Checking Requirement: Uninterpretable features must be checked (and once checked, they are deleted) Checking Requirement: Uninterpretable features must be checked (and once checked, they are deleted) Checking (under sisterhood): An uninterpretable feature F on a syntactic object Y is checked when Y is sister to another syntactic object Z which bears a matching feature F. Checking (under sisterhood): An uninterpretable feature F on a syntactic object Y is checked when Y is sister to another syntactic object Z which bears a matching feature F.

Feature checking To distinguish interpretable features from uninterpretable features, we will write uninterpretable features with a u in front of them. To distinguish interpretable features from uninterpretable features, we will write uninterpretable features with a u in front of them. D has uninterpretable feature F D has uninterpretable feature F E has interpretable feature F. E has interpretable feature F. If we Merge them, the uninterpretable feature can be checked (under sisterhood). If we Merge them, the uninterpretable feature can be checked (under sisterhood). D[uF]D[uF] E [F]

Feature checking To distinguish interpretable features from uninterpretable features, we will write uninterpretable features with a u in front of them. To distinguish interpretable features from uninterpretable features, we will write uninterpretable features with a u in front of them. D has uninterpretable feature F D has uninterpretable feature F E has interpretable feature F. E has interpretable feature F. If we Merge them, the uninterpretable feature can be checked (under sisterhood). If we Merge them, the uninterpretable feature can be checked (under sisterhood). C D[uF]D[uF] E [F]

Feature checking Or, for a more concrete example Or, for a more concrete example kick is a verb (has an interpretable V feature) and c-selects a noun (has an uninterpretable N feature). kick is a verb (has an interpretable V feature) and c-selects a noun (has an uninterpretable N feature). me is a noun (a pronoun in fact, has an interpretable N feature, and others like accusative case, first person, singular) me is a noun (a pronoun in fact, has an interpretable N feature, and others like accusative case, first person, singular) kick [uN, V] me [N, acc, 1, sg]

Feature checking Or, for a more concrete example Or, for a more concrete example kick is a verb (has an interpretable V feature) and c- selects a noun (has an uninterpretable N feature). kick is a verb (has an interpretable V feature) and c- selects a noun (has an uninterpretable N feature). me is a noun (a pronoun in fact, has an interpretable N feature, and others like accusative case, first person, singular) me is a noun (a pronoun in fact, has an interpretable N feature, and others like accusative case, first person, singular) Merging them will check the uninterpretable feature, and the structure can be interpreted. Merging them will check the uninterpretable feature, and the structure can be interpreted. V kick [uN, V] me [N, acc, 1, sg]

Feature checking The head is the “needy” one. The one that had the uninterpretable feature that was checked by Merge. The head is the “needy” one. The one that had the uninterpretable feature that was checked by Merge. The combination has the features of the verb kick and so its distribution will be like a verb’s distribution would be. The combination has the features of the verb kick and so its distribution will be like a verb’s distribution would be. Pat wants to kick me. Pat wants to kick me. Pat wants to drive. Pat wants to drive. I like to draw elephants. I like to draw elephants. *Pat wants to elephants. *Pat wants to elephants. *I like to draw kick me. *I like to draw kick me. V kick [uN, V] me [N, acc, 1, sg]

Chris glanced at Pat Pat[ ]Chris[ ] at[]glanced[ ]

                      