Computational Psycholinguistics Syllabus Introduction Overview of Computational Psycholinguistics goal areas focus of this course Computational Psycholinguistics and NLP experimental observations and computational models Human language acquisition what we learn (representation of the linguistic knowledge: competence) how we use the acquired knowledge (production, comprehension: performance) Human language processing observed behavior: general properties competence vs. performance Competence Hypothesis Sentence processing modularity incrementality ambiguity Experimental Methods Reading time studies Eye-tracking studies Neuroscientific measures Spoken comprehension and anticipation in visual environment Linking hypotheses Modularity Modularity in acquisition order of learning representation of different types of knowledge Modularity in processing interaction of different types of knowledge in language use Evidence in support of linguistic modularity Evidence against linguistic modularity Architectures and mechanisms Parsing Mechanisms Context Free Grammars Cognitive plausibility criteria incrementality memory load Top-down parsing Bottom-up parsing Left-corner parsing Comparison Handling Ambiguity in Parsing Types of ambiguity local ambiguity global ambiguity garden-path phenomena Deterministic strategies of handling ambiguity minimal attachment late closure theta attachment argument attachment Reanalysis mechanisms backtracking parallelism Probabilistic Accounts of Language Processing Experience-based models case study: relative clause attachment Maximum likelihood methods case study: TOS tagging, lexical category Probabilistic grammars eg: Jurafsky 1996 parse ranking frame preferences construction preferences Cognitive plausibility Multiple Constraint Accounts of Language Processing Experimental evidence thematic fit lexical frequency structural bias Modularity interactive activation Eg: Competitive-Integration Model experiments computational model constraint parameters Language Acquisition Learnability and innateness experimental evidence principles and parameters framework computational modeling Usage-based accounts experimental evidence verb-island hypothesis computational modeling Modularity linking syntax and semantics, innate linking rules Construction Grammars Probabilistic modeling Frequency effects Bayesian models of acquisition Neural Networks Overview neurons vs nodes, brain vs network distributed representation and processing learning and generalization Simple architectures node structure: weights and activation functions multi-layer perceptrons: training and calculating error competitive networks acquisition case study: English past tense (Rummelhart and McClelland) Dynamic architectures notion of time and context Simple Recurrent Networks processing case study: sentence processing (Elman)