Simulating Language Behavior An Introduction C ¸a˘ grı C ¸¨ oltekin c.coltekin@rug.nl Information science/Informatiekunde 2012-02-15
Tentative Plan Week Subject 1 Introduction & Organization 2 Computational simulation of language acquisition (mostly segmentation) 3 Simulation of language change/diffusion Simulation of learning pronoun reference 4 Simulation of segmentation 5 Simulation of segmentation 6 Simulation of language comprehension Simulation of acquisition of words, morphology or syntax 7 Project presentations C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 1/34
Outline Language Behavior Modeling and Simulation Language Acquisition An example simulation: segmentation Summary & Discussion C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 1/34
Language Behavior What is language behavior? We will be dealing with questions like: ◮ How does a particular aspect of language comprehension? ◮ Why some sentences are harder to comprehend than others? ◮ How do we acquire language(s)? ◮ Is there a difference between learning regular or irregular aspects of language? ◮ How do languages change in time? ◮ What are the causes of language change, and in which ways do we expect changes to occur? C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 2/34
Language Behavior What is language behavior? We will be dealing with questions like: ◮ How does a particular aspect of language comprehension? ◮ Why some sentences are harder to comprehend than others? ◮ How do we acquire language(s)? ◮ Is there a difference between learning regular or irregular aspects of language? ◮ How do languages change in time? ◮ What are the causes of language change, and in which ways do we expect changes to occur? Note that the questions are not only related to the directly observables. It relates to how human cognitive system works . C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 2/34
Modeling and Simulation What is a model? Examples of some models in science: ◮ Galilean model of solar system. ◮ Bohr model of atom. ◮ Atmospheric models used in meteorology. ◮ Scale models of cars, bridges, buildings etc. used in engineering. ◮ Animal models used in medicine. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 3/34
Modeling and Simulation Models: why and how ◮ Why do we model things at all? C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 4/34
Modeling and Simulation Models: why and how ◮ Why do we model things at all? ◮ If the model matches the reality well, we can make predictions. ◮ We learn the phenomenon better while (formally) specifying the model. ◮ Sometimes cannot study the object of interest directly. Because it is too, expensive, unethical, or unpractical to do so. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 4/34
Modeling and Simulation Models: why and how ◮ Why do we model things at all? ◮ If the model matches the reality well, we can make predictions. ◮ We learn the phenomenon better while (formally) specifying the model. ◮ Sometimes cannot study the object of interest directly. Because it is too, expensive, unethical, or unpractical to do so. ◮ Once we have the model, how do we get knowledge out of it? C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 4/34
Modeling and Simulation Models: why and how ◮ Why do we model things at all? ◮ If the model matches the reality well, we can make predictions. ◮ We learn the phenomenon better while (formally) specifying the model. ◮ Sometimes cannot study the object of interest directly. Because it is too, expensive, unethical, or unpractical to do so. ◮ Once we have the model, how do we get knowledge out of it? ◮ Study the model analytically. ◮ Run simulations. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 4/34
Modeling and Simulation Models: why and how ◮ Why do we model things at all? ◮ If the model matches the reality well, we can make predictions. ◮ We learn the phenomenon better while (formally) specifying the model. ◮ Sometimes cannot study the object of interest directly. Because it is too, expensive, unethical, or unpractical to do so. ◮ Once we have the model, how do we get knowledge out of it? ◮ Study the model analytically. ◮ Run simulations. All models are wrong, some are useful. — Box and Draper (1986, p. 424) C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 4/34
Language Acquisition
Language Acquisition The problem of language acquisition ◮ Human languages are complex (recursion, ambiguity). ◮ Children do not receive explicit instruction during language acquisition. ◮ Language acquisition by children is (arguably) fast and robust. ◮ The input to children is not enough for learning ( Poverty of Stimulus Argument ). ◮ Children do not receive input critical for learning certain phenomena. ◮ Human languages are not learnable from positive input (Gold, 1967). Negative input is not available to children. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 5/34
Language Acquisition The debate Nativism Our knowledge of language is largely determined at birth (by our genes). Contribution of environmental factors are only of secondary importance. [...] in certain fundamental respects we do not really learn language; rather, grammar grows in the mind. (Chomsky, 1980, p.134) Plato, Descartes, Chomsky, . . . Empiricism Our knowledge is primarily due to our interactions with the environment. Aristotle, Locke, . . . Taking one of these sides is common in linguistics. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 6/34
Language Acquisition Debate resolved: we are all nativists To say that “language is not innate” is to say that there is no difference between my granddaughter, a rock, and a rabbit. In other words, if you take a rock, a rabbit, and my granddaughter and put them in a community where people are talking English, they’ll all learn English. If people believe that, then they’ll believe language is not innate. If they believe that there is a difference between my granddaughter, a rabbit, and a rock, then they believe that language is innate. — Chomsky (2000, p.50), ‘The Architecture of Language’ ( emphasis mine.) C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 7/34
Language Acquisition Debate resolved: we are all empiricist The obvious conclusion is that the real answer to the question, Where the knowledge come from, is that it comes from the interaction between nature and nurture, or what has been called “epigenesis.” Genetic constraints interact with internal and external environmental influences, and they jointly give rise to the phenotype. — Elman et al. (1996, pp.i–ii), ‘Rethinking Innateness’ C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 8/34
Language Acquisition Debate in linguistics We all agree that, ◮ Part of our linguistic abilities comes from our experience: people are typically able to learn more different languages than they grow different physical organs. ◮ Part of our linguistic abilities are innate: rocks and rabbits aside, even the species closest to us cannot match with our linguistic abilities. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 9/34
Language Acquisition Debate in linguistics We all agree that, ◮ Part of our linguistic abilities comes from our experience: people are typically able to learn more different languages than they grow different physical organs. ◮ Part of our linguistic abilities are innate: rocks and rabbits aside, even the species closest to us cannot match with our linguistic abilities. The disagreement seems to be on whether the innate component is language-specific knowledge or domain-general learning abilities . C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 9/34
Language Acquisition Now we know what it is, is the debate resolved? Short answer: C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 10/34
Language Acquisition Now we know what it is, is the debate resolved? Short answer: No. ◮ It is difficult to know the quantity/type of innate knowledge necessary for settling the debate: The target seems to be moving: from P&P (Chomsky, 1981) to recursion (Hauser, Chomsky & Fitch, 2002) / merge (Berwick et al., 2011). C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 10/34
Language Acquisition Now we know what it is, is the debate resolved? Short answer: No. ◮ It is difficult to know the quantity/type of innate knowledge necessary for settling the debate: The target seems to be moving: from P&P (Chomsky, 1981) to recursion (Hauser, Chomsky & Fitch, 2002) / merge (Berwick et al., 2011). ◮ Empirical evidence is scarce, and interpreted differently. C ¸. C ¸¨ oltekin, Informatiekunde Simulating Language Behavior 10/34
Recommend
More recommend
