Psy1302 Psychology of Language Lecture 12 Sentence Comprehension II.

Slides:



Advertisements
Similar presentations
Semantics (Representing Meaning)
Advertisements

STAGES OF COMPREHENSION discourse modelling semantic analysis syntactic “parsing” lexical access phonemic analysis sensory processing.
TOWARDS A MODULAR APPROACH TO ANAPHORIC PROCESSING: semantic operations precede discourse operations Arnout Koornneef.
Eye Movements and Spoken Language Comprehension: effects of visual context on syntactic ambiguity resolution Spivey et al. (2002) Psych 526 Eun-Kyung Lee.
Sentence Processing III Language Use and Understanding Class 12.
Psycholinguistics 06.
The Interaction of Lexical and Syntactic Ambiguity by Maryellen C. MacDonald presented by Joshua Johanson.
Sentence Processing 1: Encapsulation 4/7/04 BCS 261.
Prosodic facilitation and interference in the resolution of temporary syntactic closure ambiguity Kjelgaard & Speer 1999 Kent Lee Ψ 526b 16 March 2006.
Theeraporn Ratitamkul, University of Illinois and Adele E. Goldberg, Princeton University Introduction How do young children learn verb meanings? Scene.
Language and Cognition Colombo, June 2011 Day 2 Introduction to Linguistic Theory, Part 4.
Using prosody to avoid ambiguity: Effects of speaker awareness and referential context Snedeker and Trueswell (2003) Psych 526 Eun-Kyung Lee.
9/22/10Psyc / Ling / Comm 525 Fall 10 Semantic Priming (Phenomenon & Tool)...armkitchentree Related prime >doctoractor < Unrelated prime nurse floor...
Using disfluency to understand, um, sentences... with PP-attachment ambiguities Jennifer E. Arnold and Kellen Carpenter, UNC Chapel Hill Background 1)
Statistical NLP: Lecture 3
Introduction and Jurafsky Model Resource: A Probabilistic Model of Lexical and Syntactic Access and Disambiguation, Jurafsky 1996.
SYNTAX 5 ON-LINE PROCESSING DAY 34 – NOV 15, 2013 Brain & Language LING NSCI Harry Howard Tulane University.
SYNTAX 1 DAY 30 – NOV 6, 2013 Brain & Language LING NSCI Harry Howard Tulane University.
Announcements Sample exam questions Sample exam questions –This week (Thursday): You will submit your Qs into dropbox Bring Completed Homework Bring Completed.
Language Use and Understanding BCS 261 LIN 241 PSY 261 CLASS 12: SNEDEKER ET AL.: PROSODY.
Linguistic Theory Lecture 8 Meaning and Grammar. A brief history In classical and traditional grammar not much distinction was made between grammar and.
Discourse influences during parsing are delayed Keith Rayner, Simon Garrod,& Charles A. Perfetti Cognition, 45, 1992.
1 Lexicon, experimental Oct 22, Psycholinguistic ways of examining the lexicon/syntax Three things we will look at: a. Mental Lexicon b. Collocates.
Psy1302 Psychology of Language Lecture 10 Ambiguity Resolution Sentence Processing I.
PSY 369: Psycholinguistics Language Comprehension: Sentence comprehension.
Amirkabir University of Technology Computer Engineering Faculty AILAB Efficient Parsing Ahmad Abdollahzadeh Barfouroush Aban 1381 Natural Language Processing.
1 Introduction to Computational Linguistics Eleni Miltsakaki AUTH Spring 2006-Lecture 4.
The Neural Basis of Thought and Language Week 15 The End is near...
grateful acknowledgments to
Language, Mind, and Brain by Ewa Dabrowska Chapter 2: Language processing: speed and flexibility.
Day 2: Pruning continued; begin competition models
PSY 369: Psycholinguistics Language Comprehension: The role of memory.
Understanding Sentences. Two steps back: What is linguistic knowledge? Phonological Syntactical Morphological Lexical Semantic.
PSY 369: Psycholinguistics Language Comprehension: Sentence comprehension.
The time-course of prediction in incremental sentence processing: Evidence from anticipatory eye movements Yuki Kamide, Gerry T.M. Altman, and Sarah L.
1 LIN 1310B Introduction to Linguistics Prof: Nikolay Slavkov TA: Qinghua Tang CLASS 14, Feb 27, 2007.
10/6/10Psyc / Ling / Comm 525 Fall 10 More on predicting word properties in context Dikker, Rabagliati, Farmer, & Pylkkanen (2010) Psych Science MagnetoEncephaloGraphy.
Jelena Mirković and Maryellen C. MacDonald Language and Cognitive Neuroscience Lab, University of Wisconsin-Madison Introduction How to Study Subject-Verb.
Lecture 12: 22/6/1435 Natural language processing Lecturer/ Kawther Abas 363CS – Artificial Intelligence.
Ferreira and Henderson (1990)
Speech Comprehension: Decoding meaning from speech.
Older Adults’ More Effective Use of Context: Evidence from Modification Ambiguities Robert Thornton Pomona College Method Participants: 32 young and 32.
The Independence of Syntactic Processing Advanced Psycholinguistics Presenter: Dong-Bo Hsu 02/09/06.
1 Natural Language Processing Lecture 11 Efficient Parsing Reading: James Allen NLU (Chapter 6)
3/5/08Psyc / Ling / Comm 525 Spring08 Back to Usual Issues So, evidence supports both Parallelism & Interaction of multiple within-sentence constraints.
Origins of Modularity. Preliminary Binding Study Results –13 participants tested using Linger –Data can be analyzed using Lingalyzer Lite –Region encoding.
PSY 369: Psycholinguistics Language Comprehension: The role of memory.
Modularity What’s the Big Deal? (1983) (not 1983)
PS: Introduction to Psycholinguistics Winter Term 2005/06 Instructor: Daniel Wiechmann Office hours: Mon 2-3 pm Phone:
Avoiding the Garden Path: Eye Movements in Context
Linguistic Essentials
PSY270 Michaela Porubanova. Language  a system of communication using sounds or symbols that enables us to express our feelings, thoughts, ideas, and.
Rules, Movement, Ambiguity
E BERHARD- K ARLS- U NIVERSITÄT T ÜBINGEN SFB 441 Coordinate Structures: On the Relationship between Parsing Preferences and Corpus Frequencies Ilona Steiner.
Results of Eyetracking & Self-Paced Moving Window Studies DO-Bias Verbs: The referees warned the spectators would probably get too rowdy. The referees.
A Strategy for Looking For Effects of Discourse on Sentence Comprehension Look for effects of discourse context by making sentence require something from.
Dec 11, Human Parsing Do people use probabilities for parsing?! Sentence processing Study of Human Parsing.
SYNTAX 4 NOV 16, 2015 – DAY 34 Brain & Language LING NSCI Fall 2015.
Semantic indeterminacy in object relative clauses Maryellen C. MacDonald Silvia P. Gennari.
Parafoveal Preview in Reading Burgess (1991) - Self-paced moving window reading time study - Varied window size from single to several words - Found an.
48 Item Sets (Only the results for the relative clause versions are reported here.) The professor (who was) confronted by the student was not ready for.
September 26, : Grammars and Lexicons Lori Levin.
Syntactic Priming in Sentence Comprehension (Tooley, Traxler & Swaab, 2009) Zhenghan Qi.
Revision Lecture Cognitive Science. Past papers What is the answer to the question? The answer will nearly always involve: “How amazing it is that people.
Method. Input to Learning Two groups of learners each learn one of two new Semi-Artificial Languages. Both Languages: Example sentences: glim lion bee.
Top-down processing of language -necessary due to the noisy and variable nature of the stimulus -e.g.: coarticulation -luckily, we tend to engage in categorical.
Chapter 11 Language. Some Questions to Consider How do we understand individual words, and how are words combined to create sentences? How can we understand.
48 Item Sets (Only the results for the relative clause versions are reported here.) The professor (who was) confronted by the student was not ready for.
Statistical NLP: Lecture 3
Linguistic Essentials
Presentation transcript:

Psy1302 Psychology of Language Lecture 12 Sentence Comprehension II

Today Revisit lexical and syntactic ambiguity connection Revisit lexical and syntactic ambiguity connection Critiques of F&C Critiques of F&C Remaining time – Review (Q & A) Remaining time – Review (Q & A)

pitcher + supportive context port Equibias Non-Equibias + supportive context Do you remember this experiment? Supportive Context No Supportive Context Equi-bias and Non-Equibias Nouns

New View Syntactic Ambiguity Resolution is like Lexical Ambiguity Resolution Syntactic Ambiguity Resolution is like Lexical Ambiguity Resolution Lexicalist Based Constraint Satisfaction View Lexicalist Based Constraint Satisfaction View

Lexicalist Based Constraint Satisfaction View Lexical entries have information about the syntax Lexical entries have information about the syntax Lexical information of verbs include e.g.: Lexical information of verbs include e.g.: –what kind of arguments it takes (e.g., “put” takes NP, PP) (e.g., “put” takes NP, PP) –thematic role between the verb and its argument (e.g., patient/theme, goal, etc.) (e.g., patient/theme, goal, etc.) –syntactic structures & the frequency of occurrence

Classic Ambiguous Sentence The horse raced past the barn fell. The horse raced past the barn fell.

Ambiguous Sentences Which one seems more difficult? 1. The horse raced past the barn fell. 2. The horse carried past the barn fell.

Verbs and Structural Frequencies Table from MacDonald, Pearlmutter, & Seidenberg Paper

Syntactic Structures Verbs differ in preferences (frequencies) for structures Verbs differ in preferences (frequencies) for structures Which verb you choose for your studies will matter Which verb you choose for your studies will matter Past Tense (Main Clause) Past Participle (Reduced Relative).. Equi-bias and Non-Equibias Structures

Table from MacDonald, Pearlmutter, & Seidenberg Paper

examined Non-Equibias Structures Syntactic Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” -- Thickness of the line indicates amount of activation. Various contextual information can influence the levels of activation. Various contextual information can influence the levels of activation.

examined Non-Equibias Structures Syntactic Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” Thematic information: Can X do the examining? Thematic information: Can X do the examining? –Good agent  reinforce Past Tense (Main Verb) reading Past Tense (MC) Good Agent Thematic Influence on Equi-bias and Non-Equibias Structures

examined Non-Equibias Structures Syntactic Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” Thematic information: Can X do the examining? Thematic information: Can X do the examining? –Poor agent  reinforce Past Participle (Reduced Relative) reading Past Participle (RR) Poor Agent Thematic Influence on Equi-bias and Non-Equibias Structures

examined Non-Equibias Structures Syntactic Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” Thematic information: Can X be examined? Thematic information: Can X be examined? –Poor agent & Good theme  even more strongly reinforce Past Participle (Reduced Relative) reading Past Participle (RR) Poor Agent & Good Theme. Thematic Influence on Equi-bias and Non-Equibias Structures

examined Non-Equibias Structures Syntactic Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” Discourse and referential context Discourse and referential context –E.g. 2 defendants, one who was examined by the lawyer, and one who was not. Past Participle (RR) 2 Referents. Thematic Influence on Equi-bias and Non-Equibias Structures

Critiques of F&C Experiment 1 Experiment 1 –Trueswell, Tanenhaus, & Garnsey (1994) Thematic (Semantic) Information and Reduced Relative Reading Thematic (Semantic) Information and Reduced Relative Reading Experiment 2 & 3 Experiment 2 & 3 –Tabossi, Spivey-Knowlton, McRae, & Tanenhaus (1994) –Britt, Perfetti, Garrod, & Rayner (1992) Main Clause vs. Reduced Relative Main Clause vs. Reduced Relative VP-attached vs. NP-attached VP-attached vs. NP-attached

Ferreira & Clifton (1986) Q: Is the initial syntactic processing stage influenced by: 1. thematic/semantic information (Exp. 1) 2. pragmatic or contextual information (Exp. 2 & 3)

Ferreira & Clifton (1986) Experiment 1 4 Sentence Types: Reduced, Animate Reduced, Animate The defendant examined by the lawyer turned out to be unreliable. Reduced, Inanimate Reduced, Inanimate The evidence examined by the lawyer turned out to be unreliable. Unreduced, Animate Unreduced, Animate The defendant that was examined by the lawyer turned out to be unreliable. Unreduced, Inanimate Unreduced, Inanimate The evidence that was examined by the lawyer turned out to be unreliable. DISAMBIGUATING REGION

The evidence examined by the lawyer turned out to be unreliable. C C-1 C-2 C+1C+2 Animate Reduced Animate Unreduced Inanimate Reduced Inanimate Unreduced SLOW FAST SLOW FAST SLOW FAST MODULAR INTERACTIVE FAST OR SLOW PREDICTIONS ACTUAL RESULTS

Trueswell, Tanenhaus, & Garnsey (1994) Critique of Ferreira & Clifton Critique of Ferreira & Clifton 8 out of 16 of Ferreira & Clifton’s INANIMATE items had possible main clause continuations. 8 out of 16 of Ferreira & Clifton’s INANIMATE items had possible main clause continuations.

Examples from F&C Continue with non-relative clause reading? The car towed… The car towed… The ship sighted… The ship sighted… The message recorded… The message recorded… The skin felt… The skin felt… The trash smelled… The trash smelled… The car sold… The car sold… Baldwin screaming at his child. Cape Cod on November 19, 1620 the truck. soft. nasty. for a million dollars

Trueswell, Tanenhaus, & Garnsey (1994) Experiment 1 Stimuli Experimenter made sure inanimate sentences were poor agents Table 1 from paper

Trueswell, Tanenhaus, & Garnsey (1994) Experiment 1 First Pass Reading

Trueswell, Tanenhaus, & Garnsey (1994) Experiment 1 Second Pass Reading

Trueswell, Tanenhaus, & Garnsey (1994) Experiment 2 Manipulations Used normed N-V stimuli Used normed N-V stimuli –Animates had 100% Main Clause sentence completion –Inanimates had under 30% Main Clause sentence completion Decreased the number of relative clause sentences tested Decreased the number of relative clause sentences tested –Decrease chance of sentence structure becoming expected or priming other RR sentences Used mixed case instead of all upper case. Used mixed case instead of all upper case. –Increased naturalness and ease of text read –Reduce reading time  reduce processing/ambiguity resolution time

Trueswell, Tanenhaus, & Garnsey (1994)

Inanimate stimuli were “bad agents” Inanimate stimuli were “bad agents” Were the stimuli “good patient/theme”? Were the stimuli “good patient/theme”? Rating Task: Rating Task: –“How typical is it for the evidence to examine something?” –“How typical is it for the evidence to be examined by someone?” 1 = not typical, 7 = typical

Trueswell, Tanenhaus, & Garnsey (1994) p <.05 = statistically significant High Rating (good theme/patient) = Short Reading time (easier processing)

Trueswell, Tanenhaus, & Garnsey (1994) Strong Semantic fit defined as: Agent rating 5.0 Weak Semantic Fit: remaining items. Y-axis Reading Time Difference = Reduced minus Unreduced Relative Clause Weak fit Inanimates Strong Fit Inanimates Unambiguous (e.g. drawn) examined by the lawyer Weak fit Animates examined by the lawyer

Syntactic Structures Thematic information: Can X do the examining? Thematic information: Can X do the examining? + Good agent  reinforce Past Tense (Main Verb) reading + Poor agent  reinforce Past Participle (Reduced Relative) reading + Poor agent & Good theme  even more strongly reinforce Past Participle (Reduced Relative) reading examined Non-Equibias Structures Past Tense (MC) Past Participle (RR) “The defendant/evidence examined….” F&C’s Inanimates PT (MC) PP (RR)

examinedby the lawyer turned out Reading Time Animate Reduced Animate Unreduced Inanimate Reduced Inanimate Unreduced examinedby the lawyer turned out Animate Reduced Animate Unreduced Inanimate Reduced Inanimate Unreduced F&C: Exp. 1 Data Revisited

Ferreira & Clifton (1986) Q: Is the initial syntactic processing stage influenced by: 1. thematic/semantic information (Exp. 1) 2. pragmatic or contextual information (Exp. 2 & 3) 1.What was tested? 2.Critiques?

Main Clause vs. Reduced Relative Main Clause frequency is 12x that of Reduced Relative Main Clause frequency is 12x that of Reduced Relative Past Tense (Main Clause) Past Participle (Reduced Relative) Context has to be really strong to overcome Main Clause reading! Context has to be really strong to overcome Main Clause reading!

Complete the sentence #1 John worked as a reporter for a big city newspaper. He sensed that a major story was brewing over the city hall scandal, and he obtained some evidence that he believed pretty much established the mayor’s guilt. He went to his editors with a tape and some photos because he needed their approval before he could go ahead with the story. He ran a tape for one of his editors, and he showed some photos to the other. The editor played the tape… Tally up homework survey

Complete the sentence #1 how did your friend complete the sentence/paragraph? “played the tape”… –Main Clause –Reduced Relative Clause –Other NP VP S NP V CONJ VP NP V The editor playedthe tape andlikedit MAIN CLAUSE REDUCED RELATIVE CLAUSE NP VP S NP V VP NP V The man played the tapelikedit NP t S

Main Clause vs. Reduced Relative F&C’s context is weak F&C’s context is weak –Context is weak as determined by other researchers using sentence completion at point of ambiguity. –Context is insufficient to overcome preference for reduced relative clause reading Context might just be enough to make relative clause in competition with the dominant main verb reading  Increased reading times Context might just be enough to make relative clause in competition with the dominant main verb reading  Increased reading times –(F&C Exp 3)

F&C: Experiment 2 & 3 Referential Context Information Support + NMA Support + MA No Support + MA No Support + NMA

F&C: Experiment 3 Faster With context. Slower With context. Context x Attachment Effect!!!

F&C: Experiment 3 Replication of Experiment 2 with another method Context supporting Minimal Attachment Reduces Reading Time Context supporting Minimal Attachment Reduces Reading Time Context supporting Non-Minimal Attachment Increases Reading Time Context supporting Non-Minimal Attachment Increases Reading Time Why? Why? PT (MC) PP (RR) +MA supportive context MA sentence NMA sentence PT (MC) PP (RR) +NMA supportive context PT (MC) PP (RR) PT (MC) PP (RR)

VP-attached vs. NP- attached Last lecture, we saw a case where referential context affects VP vs. NP attachment preference. Last lecture, we saw a case where referential context affects VP vs. NP attachment preference. What’s going on in F&C’s study? What’s going on in F&C’s study?

Complete the sentence #2 Sam worked at a factory warehouse. His job was to make sure that boxes of merchandise were ready to be delivered. Sam had to fill up a van so it could go out. He had a pile of boxes on a cart and another pile on the floor. He knew some guys from another department needed the cart. Sam loaded the boxes on the cart… Tally up homework survey

Complete the sentence #2 “on the cart”: –VP-attached –NP-attached –Other NP PP S VP NP V Sam the boxes PP loaded on the cart before lunch S VP NP V PP onto the van Sam the boxes loadedon the cart VP-attached NP-attached

VP-attached vs. NP- attached PP VP-Attached PP NP-Attached PUT (V): NP, PP

VP-attached vs. NP- attached Lexical Biases of Verbs in F&C Lexical Biases of Verbs in F&C –Verbs highly supportive of VP-attached reading over NP-attached reading E.g. “load”, “place” expect PP. E.g. “load”, “place” expect PP. “ Sam loaded the boxes on the cart. ” “ Sam loaded the boxes on the cart. ” –VP-attached frequent

VP-attached vs. NP- attached What if we used other verbs? (Britt et al. 1992) What if we used other verbs? (Britt et al. 1992) Peter read the books on the chair instead of lying in bed (VP-attachment) Peter read the books on the chair instead of the other books (NP-attachment) Peter read the books on the war instead of the other books (NP-attachment) vs. 3 in neutral context. VP-attachment is read faster 1 vs. 3 in supportive context. Both equally fast. 1 vs. 2 in neutral context. VP-attachment is faster. 1 vs. 2 in supportive context. Both equally fast. Think about analogy to lexical ambiguities!

Models of Sentence Processing Garden-Path Model Garden-Path Model –Autonomous Late closure Late closure Minimal attachment Minimal attachment Constraint-Based Model Constraint-Based Model –Interactive Lexical Biases Lexical Biases Referential Contexts Referential Contexts Structural Biases Structural Biases } Cues from multiple sources constrain interpretation

Final Word of Caution (a quote from MacDonald, Pearlmutter, & Seidenberg, 1994) [A]lthough the architecture… affords the possibility of continuous interaction between contextual information and the lexicon, the effects of contexts tend to be more retroactive than proactive…. [A]lthough the architecture… affords the possibility of continuous interaction between contextual information and the lexicon, the effects of contexts tend to be more retroactive than proactive…. [L]ess information is needed to discriminate between two alternatives than to preselect one of them. [L]ess information is needed to discriminate between two alternatives than to preselect one of them.