Modeling Reading, Spelling, and Past Tense Learning with Artificial Neural Networks

Bullinaria, John A.

doi:10.1006/brln.1997.1818

Cited by 64 publications

(49 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Then Plaut et al [14] developed improved models, but returned to pre-aligned training data to get good performance. This led Bullinaria to explore extensions of the NETtalk model that were able to learn appropriate alignments at the same time as learning the mapping process [3,5,6].…”

Section: The Basic Neural Networkmentioning

confidence: 99%

“…This basic architecture and learning algorithm formed the basis of a series of models of reading aloud, including human-like generalization ability for reading non-words, accounts of frequency and regularity effects in reaction times, and models of developmental and acquired surface dyslexia [3,6]. Relatively straightforward extensions of it were also used to model the harder reverse task of spelling, i.e.…”

Section: The Basic Neural Networkmentioning

confidence: 99%

“…Relatively straightforward extensions of it were also used to model the harder reverse task of spelling, i.e. phoneme-to-text conversion [4,6]. One might wonder, therefore, why this process has not been applied to text-tophoneme mapping more generally.…”

Section: The Basic Neural Networkmentioning

confidence: 99%

“…One reason is that most of the older neural network models [3,6,14,16] were aimed mainly at modeling psychological data and understanding human language abilities, rather than producing high performance applications for speech technology. They therefore concentrated on producing human-like performance on small-scale empirically testable data-sets rather than large-scale systems useable for real world applications.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, the computational resources required for training such neural networks has prevented scaling them up to larger data-sets. However, computers are now much more powerful, and a simple scalable neural network based approach for dealing with both the alignment and mapping problems has existed in the psychological modelling literature for some time [3,4,5,6], so it is worth exploring what can now be achieved with a neural network approach to this problem. …”

mentioning

confidence: 99%

See 4 more Smart Citations

Text to phoneme alignment and mapping for speech technology: A neural networks approach

Bullinaria

2011

The 2011 International Joint Conference on Neural Networks

Self Cite

View full text Add to dashboard Cite

Abstract-A common problem in speech technology is the alignment of representations of text and phonemes, and the learning of a mapping between them that generalizes well to unseen inputs. The state-of-the-art technology appears to be symbolic rule-based systems, which is surprising given the number of neural network systems for text to phoneme mapping that have been developed over the years. This paper explores why that may be the case, and demonstrates that it is possible for neural networks to simultaneously perform text to phoneme alignment and mapping with performance levels at least comparable to the best existing systems. I. INTRODUCTIONMany speech technology applications rely on having a good mapping from text to phonemes that not only performs perfectly on known words, but also generalizes well to new words, such as previously unseen proper nouns [8].Alignment of the text and phonemes is the first stage of data processing necessary to provide useable training data for many text to phoneme conversion systems, including the most successful symbolic rule-based systems [8,10] and most neural network systems [14,16,17]. The state-of-theart for this alignment process appears to be the rule-based Expectation-Maximization (EM) algorithm of Damper et al. [9], and data aligned in that way has been employed in the rule-based Pronunciation by Analogy (PbA) system of Damper et al. [8,9] to provide state-of-the-art text to phoneme mappings for English [10].Given the number of neural network based systems that have been developed over the years, it seems surprising that they are not more competitive in this area. One reason is that most of the older neural network models [3,6,14,16] were aimed mainly at modeling psychological data and understanding human language abilities, rather than producing high performance applications for speech technology. They therefore concentrated on producing human-like performance on small-scale empirically testable data-sets rather than large-scale systems useable for real world applications. Moreover, the computational resources required for training such neural networks has prevented scaling them up to larger data-sets. However, computers are now much more powerful, and a simple scalable neural network based approach for dealing with both the alignment and mapping problems has existed in the psychological modelling literature for some time [3,4,5,6], so it is worth exploring what can now be achieved with a neural network approach to this problem.

show abstract

Section: The Basic Neural Networkmentioning

confidence: 99%

Section: The Basic Neural Networkmentioning

confidence: 99%

Section: The Basic Neural Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Text to phoneme alignment and mapping for speech technology: A neural networks approach

Bullinaria

2011

The 2011 International Joint Conference on Neural Networks

Self Cite

View full text Add to dashboard Cite

show abstract

Connectionist Approaches to Reading

Plaut

The Science of Reading: A Handbook

View full text Add to dashboard Cite

Reading is a highly complex task involving the rapid coordination of visual, phonological, semantic, and linguistic processes. Computational models have played a key role in the scientific study of reading. These models allow us to explore the implications of specific hypotheses concerning the representations and processes underlying reading acquisition and performance. A particular form of computational modeling, known as connectionist or neural network modeling, offers the further advantage of being explicit about how such mechanisms might be implemented in the brain.In connectionist models, cognitive processes take the form of cooperative and competitive interactions among large numbers of simple neuron-like processing units. Typically, each unit has a real-valued activity level, roughly analogous to the firing rate of a neuron. Unit interactions are governed by weighted connections that encode the longterm knowledge of the system and are learned gradually through experience. Units are often organized into layers or groups; the activity of some groups of units encode the input to the system; the resulting activity of other groups of units encodes the system's response to that input. For example, one group might encode the written form (orthography) of a word, another might encode its spoken form (phonology), and a third might encode its meaning (semantics; see figure 2.1). The patterns of activity of the remaining groups of units -sometimes termed "hidden" units -constitute learned, internal representations that mediate between inputs and outputs. In this way, the connectionist approach attempts to capture the essential computational properties of the vast ensembles of real neuronal elements found in the brain using simulations of smaller networks of more abstract units. By linking neural computation to behavior, the framework enables developmental, cognitive, and neurobiological issues to be addressed within a single, integrated formalism. One very important advantage of connectionist models is that they deal explicitly with learning. Though many of these models have focused predominantly on simulating aspects of adult, rather than children's, reading, many of the models do explicitly consider the process of learning (e.g.,

show abstract

Modeling Typical and Atypical Cognitive Development: Computational Constraints on Mechanisms of Change

Thomas¹,

Karmiloff‐Smith²

2002

Blackwell Handbook of Childhood Cognitive Development

View full text Add to dashboard Cite

Modeling Reading, Spelling, and Past Tense Learning with Artificial Neural Networks

Cited by 64 publications

References 34 publications

Text to phoneme alignment and mapping for speech technology: A neural networks approach

Text to phoneme alignment and mapping for speech technology: A neural networks approach

Connectionist Approaches to Reading

Modeling Typical and Atypical Cognitive Development: Computational Constraints on Mechanisms of Change

Contact Info

Product

Resources

About