Robert Berwick. The Acquisition of Syntactic Knowledge. MIT Press, Cambridge, MA, 1985.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....can observe the context and guess what the speaker might have meant. In fact, many models of language acquisition assume that the input the child receives consists not only of the sentence, but also of a representation of the meaning of the sentence ( Wexler and Culicover 1980] Pinker 1984] [Berwick 1985], Briscoe 1999] among others) How much of the context surrounding the hearing of a sentence is used by children is an open question. Besides, as Landau and Gleitman [1985] observe, blind children have a severely limited access to non linguistic aspects of the world, but they succeed in learning ....

.... through Chomsky s model of a child searching through a restricted hypothesis space of grammars in order to select the correct one [Chomsky 1965] Constraints help the learner get closer to the target hypothesis by excluding many incorrect hypotheses that do not conform to these constraints [Berwick 1985]. The UG can then be considered a restriction in the hypothesis space, such that it only contains those grammars that conform to the principles and to the possible assignments of values of the parameters. In this way, a child does not need to consider every possible grammar, but only the ones ....

Berwick, R. The Acquisition of Syntactic Knowledge. MIT Press, 1985.

....acquisition device to guide its search for the target language. 6.1. Generative Framework. According to modern generative linguistics, language acquisition is the process of deterministically fixing the values of a finite set of parameters upon exposure to a finite set of linguistic evidence [3, 1, 8]. In the generative framework, a metrical theory consists of (i) a fixed set of n finite valued parameters = 1 ; 2 ; n that represent the humanly possible metrical systems; and (ii) a function m that maps a syllable structure oe and a of parameter settings to metrical grid g ....

Robert Berwick, The acquisition of syntactic knowledge, MIT Press, Cambridge, 1985.

....supported by IE approches. This area is clearly a cross road of several research lines including natural language processing (NLP) machine learning (ML) as well as software engineering (SE) Early experiences of ML for NL include works on computational learnability of linguistic knowledge ([4, 5]) or example driven learning applied to text classification and document processing ( 14] More recent approaches span from very specific language processing tasks (e.g. induction of PP disambiguation rules [8] ML driven subcategorization frames acquisition [2] or case based discourse modeling ....

R. C. Berwick, The Acquisition of Syntactic Knowledge, MIT Press, Cambridge, MA, 1985.

....learning times Table 2 shows the stress systems grouped by their theoretical analyses in terms of the parameter scheme discussed in Section 2.4. The last column of the table shows the average learning time in epochs for each group of stress patterns 8 . 7 As an example, the Subset Principle ( Berwick 85] Wexler 87] has implications for markedness. Suppose that two possible settings a and b for parameter P result in the learner respectively accepting sets S a and S b of linguistic forms. If S a is a subset of S b , then, once P has been set to value b, no positive evidence can ever re set it ....

Berwick, R., The Acquisition of Syntactic Knowledge, MIT Press, Cambridge, MA, 1985.

....learning function maps the input data to that value of a parameter which generates a language: a) compatible with the input data; and (b) smallest among the languages compatible with the input data. Briefly, the Subset Principle is a learning principle that has been proposed, originally by Berwick (1985) and refined by Manzini and Wexler, to prevent language learners from getting into a situation where negative evidence is needed in order to converge on the target grammar. Such a situation can arise if the learner somehow adopts a grammar which is a superset of the target grammar, perhaps through ....

Berwick, Robert (1985). The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.

....PII: S0893 6080(98)00139 7 otherwise cope with conditions of sparse input. However, these algorithms typically require that words in the input (or training) corpus be labelled or tagged in some way. For example, the algorithms presented in Wexler and Culicover (1980) Pinker (1984) and Berwick (1985) all require that lexical items be tagged with a syntactic role (e.g. as in Berwick) or with an argument position (relative to the verb, as in Pinker) This tagging approach is entirely reasonable provided we bear in mind that the algorithms in question are incomplete they presuppose some ....

Berwick, R. C. (1985). The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.

.... generate (sentence correction) for example [22, 52, 29, 17, 40] and those approaches that treat cases of undergeneration as well formed sentences that the grammar cannot initially generate and try to correct the grammar such that it can generate those sentences (grammar learning) for example [3, 4, 51, 47, 5, 8, 32, 7]) Sentence correction is only a temporary solution to the problem of undergeneration. Grammar learning, on the other hand, is permanent. Consequently, most contemporary work uses grammar learning as the method of overcoming undergeneration. Note that in some cases, it might be more appropriate to ....

....follows from the generally accepted fact that natural languages are infinite. Clearly, any training set will be finite and hence the learner will need to deal with the resulting sparse statistics [19] 2. 2 Model based learning The other approach to grammar learning is model based (for example [4, 53]) Model based (or deductive) learners are far less frequently used in NLP systems than data driven approaches. However, they are used by language acquisition theorists (for example [53] Roughly speaking, these methods consist of formulating general concepts from specific training examples. ....

[Article contains additional citation context not shown here]

Robert C. Berwick. The acquisition of syntactic knowledge. MIT Press, 1985.

....scientific disciplines, and there is a voluminous literature. Mathematical and computational theorists are concerned with the basic questions and definitions of language learning (Gold, 1967) with understanding the complexity of the problem (Angluin, 1978; Gold, 1978) or with good algorithms (Berwick, 1985; Rivest Schapire, 1987) An excellent survey of this approach to the problem has been written by (Angluin Smith, 1983) Linguists are concerned with grammatical frameworks which can adequately explain the basic fact that children acquire their language (Chomsky, 1965; Wexler Culicover, ....

Berwick, R. (1985). The Acquisition of Syntactic Knowledge. Cambridge: MIT Press.

....be no positive evidence to lead the learner back to the subset value. The learner will be stuck with an overgeneral grammar. Since first language acquisition, under normal circumstances, is successful, the learner must be able to avoid this subset learning problem somehow. The Subset Principle (Berwick, 1985; Manzini and Wexler, 1987; Wexler and Manzini, 1987) has been postulated in an attempt to solve this problem, in effect, by circumventing it altogether. According to the Subset Principle, the learner faced with a learning situation such as (2) is forced to choose the subset value (when both ....

....is said to instantiate the [ strict adjacency] value of the Case Adjacency parameter, while French intantiates [ strict adjacency] 7) a. Marie fait lentement ses devoirs. b. Mary does slowly her homework. White suggests that this parameter presents a subset learning problem (see also Berwick, 1985), as the [ strict adjacency] option only allows sentences observing adjacency, while the [ strict adjacency] option allows sentences that either observe or violate adjacency. More concretely, the subset language, e.g. English, allows Verb Object sequences; while the superset language, e.g. ....

Berwick, R. (1985). The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.

....stream. For example, in the sentences: 1 Sam chases the cat 2 The cat ran down the road a determiner precedes a noun three times, and hence constitutes a regularity. Such regularities can then be recorded to form a grammar. The other approach at grammar learning is modelbased (for example (Berwick 1985)) Model based (or deductive) learners are far less frequently used in NLP systems than data driven approaches. However, they are used by language acquisition theorists (for example (White 1989) Roughly speaking, these methods consist of trying to determine if a specific training example ....

.... an object grammar could be produced by computing the deductive closure of the model (e.g. a meta rule can be applied to those object rules that account for active sentences to produce object rules for passive sentences) An example of purely model based language learning is given by Berwick (Berwick 1985). More usually, though, the model is incomplete and this leads us to give it a different role in our architecture. Our model currently consists of GPSG Linear Precedence (LP) rules (Gadzar et al. 1985) semantic types (Casadio 1988) a Head Feature Convention (Gadzar et al. 1985) and X bar syntax ....

Berwick, R. C. 1985. The acquisition of syntactic knowledge. MIT Press.

....including text parsing. The chapter now goes on to present a series of grammar learners. CHAPTER 2. RELATED WORK 26 2.3. 1 Berwick Robert Berwick bases his grammar acquisition work on Chomskyan ideas of Universal Grammar (which uses a model based informant) and on Marcus Parsifal parser [77] [4], which is (generally held to be) deterministic. Parsifal consists of two data structures: a pushdown stack and a three cell lookahead buffer. The first cell may be filled with either a word or a phrasal category and is the item under consideration when parsing. Acquisition is simple. If the ....

Robert C. Berwick. The acquisition of syntactic knowledge. MIT Press, 1985.

....and Kegl et al. (1999) argue that the first learners of Nicaraguan Sign Language were such a community. Possibly, the interplay of pidgin data with inductive bias is more complex than my simulations allow, or language learning is more inventive than a strict interpretation of the Subset Principle (Berwick, 1985) suggests 7 Conclusions See Slide 14 for main focussed conclusions below some meanderings: For some types of language change the idealisation of the dynamical model, D, to infinite populations may not be harmful; for example, lexical) diffusion through American English within the last 50 ....

Berwick, Robert (1985) The Acquisition of Syntactic Knowledge, Cambridge MA: MIT Press.

.... computer program [77, 50] and Gold s model of language learning from text (positive information) by machine has been very influential in contemporary theories of natural language and in mathematical work explicitly motivated by its possible connection to human language learning (see, for example, [76, 93, 94, 66, 68, 8, 44, 15, 69, 70, 38, 39, 53, 5]) In the present paper we consider some new criteria of success extending Gold s basic model above. Suppose that we fix an integer n 0. Consider the following criterion of success (again based on (1.1) above) We say that M TxtFex n identifies L def # M, on any text for L, outputs ....

....a class of languages L, then for each L # L there is a finite set D (called a tell tale) contained in L such that D is not contained in any proper sublanguage of L in L. Intuitively, this necessary condition prevents overgeneralization in 1944 JOHN CASE learning from positive data [1, 8]. Theorem 4.4 below generalizes the subset principle to the criteria of success TxtFex a b identification and TxtBc a identification, where the a in TxtBc a identification allows each of the infinitely many final grammars converged to have up to a anomalies. Theorem 4.4 is also used to ....

[Article contains additional citation context not shown here]

R. Berwick, The Acquisition of Syntactic Knowledge, MIT Press, Cambridge, MA, 1985.

No context found.

Robert Berwick. The Acquisition of Syntactic Knowledge. MIT Press, Cambridge, MA, 1985.

No context found.

Robert C. Berwick, The Acquisition of Syntactic Knowledge, MIT Press, 1985.

No context found.

Berwick, R.C., "The Acquisition of Syntactic Knowledge", Cambridge MIT Press, 1985.

No context found.

Berwick, R. 1985. The Acquisition of Syntactic Knowledge. Cambridge, Mass.: MIT Press.

No context found.

R. Berwick, The Acquisition of Syntactic Knowledge, MIT Press, Cambridge, Mass., 1985.

No context found.

Berwick R. C., The Acquisition of Syntactic Knowledge, The MIT Press, Cambridge, Massachusetts, London, England, 1985.

No context found.

Science 274: 5294. Berwick, R. (1985). The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.

No context found.

R. C. Berwick. The Acquisition of Syntactic Knowledge. The MIT Press, Cambridge, MA, 1985.

No context found.

Robert C. Berwick. 1985. The acquisition of syntactic knowledge. MIT press, Cambridge.

No context found.

R. C. Berwick. The Acquisition of Syntactic Knowledge. MIT Press, 1985.

No context found.

Berwick, Robert (1985) The Acquisition of Syntactic Knowledge, Cambridge MA: MIT Press.

No context found.

Berwick, Robert. 1986. The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC