1-3 of 3 Results

  • Keywords: learning theory x
Clear all

Article

Kodi Weatherholtz and T. Florian Jaeger

The seeming ease with which we usually understand each other belies the complexity of the processes that underlie speech perception. One of the biggest computational challenges is that different talkers realize the same speech categories (e.g., /p/) in physically different ways. We review the mixture of processes that enable robust speech understanding across talkers despite this lack of invariance. These processes range from automatic pre-speech adjustments of the distribution of energy over acoustic frequencies (normalization) to implicit statistical learning of talker-specific properties (adaptation, perceptual recalibration) to the generalization of these patterns across groups of talkers (e.g., gender differences).

Article

Yarden Kedar

A fundamental question in epistemological philosophy is whether reason may be based on a priori knowledge—that is, knowledge that precedes and which is independent of experience. In modern science, the concept of innateness has been associated with particular behaviors and types of knowledge, which supposedly have been present in the organism since birth (in fact, since fertilization)—prior to any sensory experience with the environment. This line of investigation has been traditionally linked to two general types of qualities: the first consists of instinctive and inflexible reflexes, traits, and behaviors, which are apparent in survival, mating, and rearing activities. The other relates to language and cognition, with certain concepts, ideas, propositions, and particular ways of mental computation suggested to be part of one’s biological make-up. While both these types of innatism have a long history (e.g., debate by Plato and Descartes), some bias appears to exist in favor of claims for inherent behavioral traits, which are typically accepted when satisfactory empirical evidence is provided. One famous example is Lorenz’s demonstration of imprinting, a natural phenomenon that obeys a predetermined mechanism and schedule (incubator-hatched goslings imprinted on Lorenz’s boots, the first moving object they encountered). Likewise, there seems to be little controversy in regard to predetermined ways of organizing sensory information, as is the case with the detection and classification of shapes and colors by the mind. In contrast, the idea that certain types of abstract knowledge may be part of an organism’s biological endowment (i.e., not learned) is typically met with a greater sense of skepticism. The most influential and controversial claim for such innate knowledge in modern science is Chomsky’s nativist theory of Universal Grammar in language, which aims to define the extent to which human languages can vary; and the famous Argument from the Poverty of the Stimulus. The main Chomskyan hypothesis is that all human beings share a preprogrammed linguistic infrastructure consisting of a finite set of general principles, which can generate (through combination or transformation) an infinite number of (only) grammatical sentences. Thus, the innate grammatical system constrains and structures the acquisition and use of all natural languages.

Article

A computational learner needs three things: Data to learn from, a class of representations to acquire, and a way to get from one to the other. Language acquisition is a very particular learning setting that can be defined in terms of the input (the child’s early linguistic experience) and the output (a grammar capable of generating a language very similar to the input). The input is infamously impoverished. As it relates to morphology, the vast majority of potential forms are never attested in the input, and those that are attested follow an extremely skewed frequency distribution. Learners nevertheless manage to acquire most details of their native morphologies after only a few years of input. That said, acquisition is not instantaneous nor is it error-free. Children do make mistakes, and they do so in predictable ways which provide insights into their grammars and learning processes. The most elucidating computational model of morphology learning from the perspective of a linguist is one that learns morphology like a child does, that is, on child-like input and along a child-like developmental path. This article focuses on clarifying those aspects of morphology acquisition that should go into such an elucidating a computational model. Section 1 describes the input with a focus on child-directed speech corpora and input sparsity. Section 2 discusses representations with focuses on productivity, developmental paths, and formal learnability. Section 3 surveys the range of learning tasks that guide research in computational linguistics and NLP with special focus on how they relate to the acquisition setting. The conclusion in Section 4 presents a summary of morphology acquisition as a learning problem with Table 4 highlighting the key takeaways of this article.