Learning to segment speech using multiple cues a connectionist model (2)
➤ Gửi thông báo lỗi ⚠️ Báo cáo tài liệu vi phạmNội dung chi tiết: Learning to segment speech using multiple cues a connectionist model (2)
Learning to segment speech using multiple cues a connectionist model (2)
LANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)en, and Mark s. SeidenbergProgram in Neural, Informational, and Behavioral Sciences, University of Southern California, USAConsiderable research in language acquisition has addressed the extent to which basic aspects of linguistic structure might be identified on the basis of probabilistic cues in c Learning to segment speech using multiple cues a connectionist model (2)aregiver speech to children. This type of learning mechanism presents classic learnability issues: there are aspects of language for which the input iLearning to segment speech using multiple cues a connectionist model (2)
s thought to provide no evidence, and the evidence that does exist tends to be unreliable. We address these issues in the context of the specific probLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)vided with information about phonemes, relative lexical stress, and boundaries between utterances. Individually these sources of information provide relatively unreliable cues to word boundaries and no direct evidence about actual word boundaries. After training on a large corpus of child-directed s Learning to segment speech using multiple cues a connectionist model (2)peech, the model was able to use these cues to reliably identify word boundaries. The model shows that aspects of linguistic structure that are not ovLearning to segment speech using multiple cues a connectionist model (2)
ertly marked in the input can be derived by efficiently combining multiple probabilistic cues. Connectionist networks provide a plausible mechanism foLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)n Neural, Informational, and Behavioral Sciences, University of Southern California, University Park MC-2520, Los Angeles. CA 90089-2520 USA.We would like to thank Michael Brent. Gary Cottrell. James Morgan, and two anonymous reviewers for their comments on a previous version of this article. We are Learning to segment speech using multiple cues a connectionist model (2) also grateful to Michael Brent for providing access to a computational implementation of the DR algorithm. The research presented here is partially sLearning to segment speech using multiple cues a connectionist model (2)
upported by NIMH grant 47566 and an NIMH Research Scientist Development Award lo MSS.222 CHRISTIANSEN, ALLEN. SEIDENBERGINTRODUCTIONIn recent years thLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)n. There is particular interest in the idea that the child’s entry into language—the initial identification of relevant phonological and lexical units and basic syntactic information such as grammatical categories—is driven by analyses of the statistical properties of input. Such properties include Learning to segment speech using multiple cues a connectionist model (2)facts about the distributions of words in contexts and correlations among different types of linguistic information. These properties of language haveLearning to segment speech using multiple cues a connectionist model (2)
been largely excluded from investigations of grammatical competence and language acquisition since Chomsky (1957).Several factors have conspired to mLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)ning the statistical structure of language (e.g. Aijmer & Altenberg, 1991) that have led to some impressive results in applied areas such as automatic speech recognition and text tagging (Brill, 1993; Church, 1987; Marcus, 1992). Second, there have been important discoveries concerning aspects of th Learning to segment speech using multiple cues a connectionist model (2)e input to the child that may provide reliable cues to linguistic structure (Echols & Newport. 1992: Jusczyk, 1993; Morgan, 1986). Third, there has beLearning to segment speech using multiple cues a connectionist model (2)
en the development of connectionist learning procedures suitable for acquiring and representing such information efficiently (Rumelhart & McClelland, LANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)ore interestingly, they are coupled with a theory of knowledge representation that permits the development of abstract, underlying structures (Elman, 1991; Hinton. McClelland, & Rumelhart, 1986). The considerable progress in these related areas provides a strong motivation for reconsidering question Learning to segment speech using multiple cues a connectionist model (2)s about language learning that many linguists assume were settled long ago.The theory that statistical properties of language play a central role in aLearning to segment speech using multiple cues a connectionist model (2)
cquisition faces two classic learnability problems. First, there is the question as to how children might learn specific aspects of linguistic structuLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)e known innately to the child because experience provides no evidence for them (Crain. 1991). A second observation about language acquisition is that the input affords unlimited opportunities to make reasonable but false generalisations, yet children rapidly converge on the grammars of the languages Learning to segment speech using multiple cues a connectionist model (2) to which they are exposed without seeming to pursue these many alternatives (Hornstein & Lightfoot, 1981). Thus, the input is said to provide both toLearning to segment speech using multiple cues a connectionist model (2)
o little evidence concerning properties of the target language and tooSEGMENTING SPEECH USING MU knowledge and constraints on learning are seen as proLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)uisition. Both of these issues are relevant to approaches to acquisition that rely on statistical properties of the input. With regard to the first claim, it must be determined whether the picture concerning “linguistic structures for which there is no evidence" changes when we consider statistical Learning to segment speech using multiple cues a connectionist model (2)properties of language that have previously been ignored. With regard to the second claim, languages can be statistically analysed in innumerable waysLearning to segment speech using multiple cues a connectionist model (2)
and therefore the problem as to how the child could know which aspects to attend to is a serious one. There is a further problem insofar as statisticLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2)re could facilitate language learning is unclear.In this article, we explore systems that are capable of learning and representing statistical properties of language such as the constellations of overlapping, partially predictive cues increasingly implicated in research on language development (e.g. Learning to segment speech using multiple cues a connectionist model (2) Morgan, & Demuth, 1996). Such cues tend to be probabilistic and violable, rather than categorical or rule-governed. Importantly, these systems incorpLearning to segment speech using multiple cues a connectionist model (2)
orate mechanisms for combining different sources of information, including cues that may not be highly constraining when considered in isolation. We eLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2) information, and that this process of integration reduces the potential for making false generalisations. Thus, the general answer we adopt to both of the classical learnability questions is that there are mechanisms for efficiently combining cues of even very low validity, that such combinations o Learning to segment speech using multiple cues a connectionist model (2)f cues are the source of evidence about aspects of linguistic structure that would be opaque to a system insensitive to such combinations, and that thLearning to segment speech using multiple cues a connectionist model (2)
ese mechanisms are used by children acquiring languages (for a similar view, see Bates & MacWhinney, 1987). These mechanisms also play a role in skillLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2) & Tanenhaus, 1994) that emphasise the use of probabilistic sources of information in the service of computing linguistic representations. Since the learners of a language grow up to use it, investigating these mechanisms provides a link between language learning and language processing. In the rema Learning to segment speech using multiple cues a connectionist model (2)inder of this article we explore these ideas as they apply to the problem of segmenting utterances into words. Although we concentrate here on the relLearning to segment speech using multiple cues a connectionist model (2)
evance of combinatorial information to this specific aspect of acquisition, our view is224 CHRISTIANSEN, ALLEN. SEIDENBERGDerived Linguistic StructureLANGUAGE AND COGNITIVE PROCESSES. 1998,13 c ,,Learning to Segment Speech Using Multiple Cues: A Connectionist ModelMorten H. Christiansen, Joseph Alle Learning to segment speech using multiple cues a connectionist model (2) an alternative view in which language acquisition can be seen as involving several simultaneous tasks. The primary task—the language learner’s goal—is to comprehend the utterances to which she is exposed for the purpose of achieving specific outcomes. In the service of this goal the child attends t Learning to segment speech using multiple cues a connectionist model (2)o the linguistic input, picking up different kinds of information, subject to perceptual and attentional constraints. There is a growing body of evideLearning to segment speech using multiple cues a connectionist model (2)
nce that as a result of attending to sequential stimuli, both adults and children incidentally encode statistically salient regularities of the signalGọi ngay
Chat zalo
Facebook