Results 1 - 10
of
10
String Extension Learning
, 2009
"... This paper defines a collection of functions which define classes of languages, which have the property that they are identifiable in the limit from positive data from a very simple kind of learner. Furthermore these learners are always incremental, maximally consistent, and locally conservative. Th ..."
Abstract
-
Cited by 12 (5 self)
- Add to MetaCart
(Show Context)
This paper defines a collection of functions which define classes of languages, which have the property that they are identifiable in the limit from positive data from a very simple kind of learner. Furthermore these learners are always incremental, maximally consistent, and locally conservative. They are also efficient provided the function itself is efficient. These learners are called string extension learners because components of the grammar are read directly from strings in the language via the defining function. A number of classes of languages in the literature can be described this way including varieties of k-Locally Testable languages (McNaughton and Papert 1971) and k-Piecewise Testable languages (Simon 1975), as well as some classes not discussed in the literature, such as the k-Piecewise Testable languages in the Strict Sense. Potential applications of string extension learning exist for models of natural languages, particularly phonotactics, aspects of cognition and natural language processing.
Learning Long-Distance Phonotactics
, 2008
"... Two questions regarding the non-local nature of long-distance agreement in consonantal harmony patterns (Hansson 2001, Rose and Walker 2004) are addressed: (1) How can such patterns be learned from surface forms alone? (2) How can we understand a a major feature of the typology—the absence of blocki ..."
Abstract
-
Cited by 12 (7 self)
- Add to MetaCart
Two questions regarding the non-local nature of long-distance agreement in consonantal harmony patterns (Hansson 2001, Rose and Walker 2004) are addressed: (1) How can such patterns be learned from surface forms alone? (2) How can we understand a a major feature of the typology—the absence of blocking effects? It is shown that a learner which generalizes only by making distinctions with respect to the order of sounds (and by not making distinctions with respect to the distance between sounds) is able to learn major classes of long-distance phonotactic patterns, and is unable to learn hypothetical long-distance phonotactic patterns with blocking effects. Thus not only is the learner able to acquire attested patterns, it explains the absence of unattested ones. Furthermore, this result lends support to the idea that long distance phonotactic patterns are phenomonologically distinct from spreading patterns contra the hypothesis of Strict Locality (Gafos 1999, et seq).
Computational Phonology – Part II: Grammars, Learning, and the Future
, 2010
"... Computational phonology studies sound patterns in the world’s languages from a computational perspective. This article shows that the similarities between different generative theories outweigh the differences, and discusses stochastic grammars and learning models within phonology from a computation ..."
Abstract
-
Cited by 3 (0 self)
- Add to MetaCart
(Show Context)
Computational phonology studies sound patterns in the world’s languages from a computational perspective. This article shows that the similarities between different generative theories outweigh the differences, and discusses stochastic grammars and learning models within phonology from a computational perspective. Also, it shows how the hypothesis that all sound patterns are subregular can be investigated, pointing the direction for future research. Taken together, these contributions show computational phonology is identifying stronger and stronger universal properties of phonological patterns, which are reflected in the grammatical formalisms phonologists employ. This article is intended primarily for phonologists who are curious about computational phonology, but do not have a rigorous background in mathematics or computation. However, it is also informative for readers with a background in computation and the basics of phonology, and who are curious about what computational analysis offers phonological theory. 1
Culminativity times harmony equals unbounded stress
, 2012
"... Well-studied computational representations of stress patterns are desirable in phonological analysis for several reasons. Perhaps the most important one is that they reveal insights which are (1) relevant to any particular theory of phonology and (2) otherwise difficult to divine. This is because th ..."
Abstract
-
Cited by 1 (1 self)
- Add to MetaCart
(Show Context)
Well-studied computational representations of stress patterns are desirable in phonological analysis for several reasons. Perhaps the most important one is that they reveal insights which are (1) relevant to any particular theory of phonology and (2) otherwise difficult to divine. This is because these representations highlight the importance
What complexity differences reveal about domains in language
, 2012
"... An important distinction between phonology and syntax has been overlooked. All phonological patterns belong to the regular region of the Chomsky Hierarchy but not all syntactic patterns do. We argue that the hypothesis that humans employ distinct learning mechanisms for phonology and syntax currentl ..."
Abstract
-
Cited by 1 (1 self)
- Add to MetaCart
(Show Context)
An important distinction between phonology and syntax has been overlooked. All phonological patterns belong to the regular region of the Chomsky Hierarchy but not all syntactic patterns do. We argue that the hypothesis that humans employ distinct learning mechanisms for phonology and syntax currently offers the best explanation for this difference. 1 A role for phonology in cognitive science Whenitcomestotheproblemofhowhumanslearnlanguage, itappearsmanycomputational learning theorists, cognitive scientists, and psychologists are primarily occupied with the problem of how humans learn to put words and morphemes together to form sentences. In this article we argue that a further understanding of how sounds are put together to form words also bears directly on fundamental questions in cognitive science. In particular, we argue that computational analysis of the typology of patterns in phonology, when compared to the typology of patterns in syntax, reveals that cognitive learning mechanisms are likely multiple and modular in nature. The skew that many researchers exhibit towards morpho-syntax may really be a skew towards studying meaning. But we believe that it is because phonological systems impose different sound patterns in different languages without contributing to meaning that they are especially interesting. That is, phonology is about “Rules without Meaning ” in Frits Staal’s (1989) terms. We also believe that an apparent lack of teleological purpose in phonology is what lessens its appeal to the outside. A good discussion of the strangeness of studying phonology is provided inKaye (1989)where he considers what a programming languagelike BASIC would The authors thank Jim Rogers for valuable discussion and an anonymous reviewer and Nick Chater for
Predicting Sequential Data with LSTMs Augmented with Strictly 2-Piecewise Input Vectors
, 2016
"... Abstract Recurrent neural networks such as Long-Short Term Memory (LSTM) are often used to learn from various kinds of time-series data, especially those that involved long-distance dependencies. We introduce a vector representation for the Strictly 2-Piecewise (SP-2) formal languages, which encode ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract Recurrent neural networks such as Long-Short Term Memory (LSTM) are often used to learn from various kinds of time-series data, especially those that involved long-distance dependencies. We introduce a vector representation for the Strictly 2-Piecewise (SP-2) formal languages, which encode certain kinds of long-distance dependencies using subsequences. These vectors are added to the LSTM architecture as an additional input. Through experiments with the problems in the SPiCe dataset
MaximumLikelihoodEstimationofFeature-basedDistributions
"... Motivated by recent work in phonotactic learning (Hayes and Wilson 2008, Albright 2009), this paper shows how to define feature-based probability distributions whose parameters can be provably efficiently estimated. The main idea is that these distributions are defined as a product of simpler distri ..."
Abstract
- Add to MetaCart
(Show Context)
Motivated by recent work in phonotactic learning (Hayes and Wilson 2008, Albright 2009), this paper shows how to define feature-based probability distributions whose parameters can be provably efficiently estimated. The main idea is that these distributions are defined as a product of simpler distributions (cf. Ghahramani and Jordan 1997). One advantage of this framework is it draws attention to what is minimally necessary to describe and learn phonological feature interactions in phonotactic patterns. The “bottom-up” approach adopted here is contrasted with the “top-down ” approach in Hayes and Wilson (2008), and it is argued that the bottom-up approach is more analytically transparent. 1
Learning Unattested Languages
"... This paper demonstrates the role of morphological alternations in learning novel phonotactic patterns. In an artificial grammar learning task, adult learners were exposed to a phonotactic pattern in which the first and last consonant agreed in voicing. Long-distance phonotactics encoded as strictly ..."
Abstract
- Add to MetaCart
(Show Context)
This paper demonstrates the role of morphological alternations in learning novel phonotactic patterns. In an artificial grammar learning task, adult learners were exposed to a phonotactic pattern in which the first and last consonant agreed in voicing. Long-distance phonotactics encoded as strictly piecewise languages suggest that first-last phonotactic patterns should be unattested in natural language. However, recent theories of morphologically induced phonological patterns predict that long-distance agreement between the first and last consonant of a word can occur when the agreement is induced as a morphological alternation. The results of two experiments support the prediction that first-last harmony patterns are more easily learned when morphological cues to the pattern are present. Participants only learned the first-last pattern when presented as a morphological alternation.
8 Culminativity times harmony equals unbounded stress
"... Well-studied computational representations of stress patterns are desirable in phonological analysis for several reasons. * Perhaps the most important one is that they reveal insights which are (i) relevant to any particular theory of phonology and (ii) otherwise difficult to divine. This is because ..."
Abstract
- Add to MetaCart
Well-studied computational representations of stress patterns are desirable in phonological analysis for several reasons. * Perhaps the most important one is that they reveal insights which are (i) relevant to any particular theory of phonology and (ii) otherwise difficult to divine. This is because these rep-
