Learning constructions and the theory of grammar

Journal of Child Language, 1980

ABSTRACTThis paper presents a hypothesis-testing theory of syntax acquisition. The first section presents our model. We claim that: (1) children learn a transformational grammar, including a set of phrase structure and transformational rules; (2) linguistic universals and Occam's razor constrain the initial hypothesis space available to the device; (3) hypotheses tested by the device consist of candidate phrase structure and transformational rules; (4) linguistic evidence confirms or disconfirms hypotheses. Specific examples of incorrect phrase structure and transformational hypotheses are presented.The second section briefly surveys other approaches to language acquisition – both syntactic and non-syntactic – and compares them to our model. In the third section, we address several methodological issues: (1) the relevance of linguistic theory to the model; (2) how the model is tested; (3) the domain of the theory.

In this first brief note (one of five), I'd like to reflect on how the Dual Mechanism Model (DMM), as compared to a Single Mechanism Model (SMM), might inform our more narrow discussion of Artificial Intelligence (AI) (discussed in Note 4), as well as inform our larger-scope discussions surrounding the 'nature of language & design' more generally. The description of our methods here will be based on the following dichotomies: [1] DMM vs. SMM (i) Whereas an SMM is solely reliant on brute-force associations which are inherently tethered to overt Learning-a frequency endeavor [+Freq], where frequency of item-based learning belongs on the vertical mode of processing (to be presented and discussed below). Such item-based learning could be thought of as 'structure-independent' since its focus is solely on the isolated item in question and not on the context of overall structure surrounding the item. (ii) Whereas a DMM is abstract and rule-based which is inherently tethered to tacit, covert Acquisition-a [-Freq] endeavor which doesn't rely on a one-one association of item, but rather can be both (i) item-based and (ii) categorical in nature, where structure-dependency is observant of category over item. Hence a DMM mode-a mode which is both 'item-based' when called upon (e.g. such as lexical learning, irregular formation over rule-based regulars, etc.) and 'categorical-based' when called upon to engage in the manipulation of symbols-is in a unique position to deliver the kind of 'learning curve' which is consistent with what we find of native language acquisition (to be presented and discussed below).

In this first brief note (one of five), I'd like to reflect on how the Dual Mechanism Model (DMM), as compared to a Single Mechanism Model (SMM), might inform our more narrow discussion of Artificial Intelligence (AI) (discussed in Note 4), as well as inform our larger-scope discussions surrounding the 'nature of language & design' more generally. The description of our methods here will be based on the following dichotomies: [1] DMM vs. SMM (i) Whereas an SMM is solely reliant on brute-force associations which are inherently tethered to overt Learning-a frequency endeavor [+Freq], where frequency of item-based learning belongs on the vertical mode of processing (to be presented and discussed below). Such item-based learning could be thought of as 'structure-independent' since its focus is solely on the isolated item in question and not on the context of overall structure surrounding the item. (ii) Whereas a DMM is abstract and rule-based which is inherently tethered to tacit, covert Acquisition-a [-Freq] endeavor which doesn't rely on a one-one association of item, but rather can be both (i) item-based and (ii) categorical in nature, where structure-dependency is observant of category over item. Hence a DMM mode-a mode which is both 'item-based' when called upon (e.g. such as lexical learning, irregular formation over rule-based regulars, etc.) and 'categorical-based' when called upon to engage in the manipulation of symbols-is in a unique position to deliver the kind of 'learning curve' which is consistent with what we find of native language acquisition (to be presented and discussed below).

Journal of Language Modelling, 2014

This paper aims to bring theoretical linguistics and cognition-general theories of learning into closer contact. I argue that linguists' notions of rich UGs are well-founded, but that cognition-general learning approaches are viable as well and that the two can and should co-exist and support each other. Specifically, I use the observation that any theory of UG provides a learning criterion -- the total memory space used to store a grammar and its encoding of the input -- that supports learning according to the principle of Minimum Description-Length. This mapping from UGs to learners maintains a minimal ontological commitment: the learner for a particular UG uses only what is already required to account for linguistic competence in adults. I suggest that such learners should be our null hypothesis regarding the child's learning mechanism, and that furthermore, the mapping from theories of UG to learners provides a framework for comparing theories of UG.

2000

This paper proposes an account of the acquisition of grammatical relations using the basic concepts of connectionism and a construction-based theory of grammar. Many previous accounts of first-language acquisition assume that grammatical relations (eg, the grammatical subject and object of a sentence) and linking rules are universal and innate; this is necessary to provide a first set of assumptions in the target language to allow deductive processes to test hypotheses and/or set parameters.