An Automated Learner for Phonology and Morphology
نویسندگان
چکیده
Linguists of many persuasions take a realist view of linguistic theory as it relates to learning: language learners, in infancy and childhood, are assumed to come equipped with whatever principles of linguistic structure and of language learning are biologically determined in our species. They encounter and process numerous input data, provided essentially at random, from the ambient language. In a few years, this processing results in mental structures—grammar, phonology, lexicon—of extraordinary scope and intricacy. Ideally, linguists seek general principles of language that guide children in their search, and sometimes they attempt to determine the particular algorithms and strategies that children might be using.
منابع مشابه
The Naturalness Differential Hypothesis: Cross-linguistic Influence and Universal Preferences in Interlanguage Phonology and Morphology
The role of the mother tongue has been a major topic of second language acquisition research over the last few decades, but despite the overwhelming empirical evidence of cross-linguistic influence in learner language a number Of questions still remain to be answered: what and how much is transferred when, how and why? This study explores the extent to which a theory of linguistic naturalness —...
متن کاملAnnotating Errors in a Hungarian Learner Corpus
We are developing and annotating a learner corpus of Hungarian, composed of student journals from three different proficiency levels written at Indiana University. Our annotation marks learner errors that are of different linguistic categories, including phonology, morphology, and syntax, but defining the annotation for an agglutinative language presents several issues. First, we must adapt an ...
متن کاملVery efficient learning of structured classes of subsequential functions from positive data
In this paper, we present a new algorithm that can identify in polynomial time and data using positive examples any class of subsequential functions that share a particular finitestate structure. While this structure is given to the learner a priori, it allows for the exact learning of partial functions, and both the time and data complexity of the algorithm are linear. We demonstrate the algor...
متن کاملTwo-level morphology as phonology: Parallel automata, simultaneous rule application, and the elsewhere condition.*
Two-level morphology is a system which relates lexical representations of morpheme sequences directly to their surface/phonetic forms, without intermediate derivational stages as in most other generative approaches. It thus draws its name from the restriction of phonology to two levels of representation, and perhaps should more properly be known as two-level phonology. The morphological compone...
متن کاملA Bayesian Evaluation of the Cost of Abstractness
In this paper, we show how the modern study of reasoning under uncertainty can shed light on problems in phonology. In particular, we show how Bayesian reasoning can be used for model comparison, modeling an ideal learner. Taking a typical case of abstractness in phonology as an example, we show that Bayesian reasoning has inherent in it a simplicity bias, which is of potentially great interest...
متن کامل