A probabilistic computational model of cross-situational word learning

Words are the essence of communication: They are the building blocks of any language. Learning the meaning of words is thus one of the most important aspects of language acquisition: Children must first learn words before they can combine them into complex utterances. Many theories have been developed to explain the impressive efficiency of young children in acquiring the vocabulary of their language, as well as the developmental patterns observed in the course of lexical acquisition. A major source of disagreement among the different theories is whether children are equipped with special mechanisms and biases for word learning, or their general cognitive abilities are adequate for the task. We present a novel computational model of early word learning to shed light on the mechanisms that might be at work in this process. The model learns word meanings as probabilistic associations between words and semantic elements, using an incremental and probabilistic learning mechanism, and drawing only on general cognitive abilities. The results presented here demonstrate that much about word meanings can be learned from naturally occurring child-directed utterances (paired with meaning representations), without using any special biases or constraints, and without any explicit developmental changes in the underlying learning mechanism. Furthermore, our model provides explanations for the occasionally contradictory child experimental data, and offers predictions for the behavior of young word learners in novel situations.     

婴幼儿词语习得中的社会性认知加工过程

有关婴幼儿词语习得的研究主要有三种理论假设:联想学习理论、制约限制理论和社会语用理论.本文将从以下四个方面着重介绍社会语用理论的相关研究:婴幼儿词语习得过程涉及哪些社会性认知加工过程,社会语用理论如何应对来自前两种理论的质疑和挑战,社会性认知加工过程的相对权重如何以及自闲症儿童词语习得存在怎样的社会性认知加工障碍.相关研究结果表明社会性认知加工过程在婴幼儿词语习得中具有基础性作用.最后对今后在该领域的研究中需要进一步深入探讨的问题提出了一些思考.     

Analyzing the Rate at Which Languages Lose the Influence of a Common Ancestor

Analyzing the rate at which languages change can clarify whether similarities across languages are solely the result of cognitive biases or might be partially due to descent from a common ancestor. To demonstrate this approach, we use a simple model of language evolution to mathematically determine how long it should take for the distribution over languages to lose the influence of a common ancestor and converge to a form that is determined by constraints on language learning. We show that modeling language learning as Bayesian inference of n binary parameters or the ordering of n constraints results in convergence in a number of generations that is on the order of n log n. We relax some of the simplifying assumptions of this model to explore how different assumptions about language evolution affect predictions about the time to convergence; in general, convergence time increases as the model becomes more realistic. This allows us to characterize the assumptions about language learning (given the models that we consider) that are sufficient for convergence to have taken place on a timescale that is consistent with the origin of human languages. These results clearly identify the consequences of a set of simple models of language evolution and show how analysis of convergence rates provides a tool that can be used to explore questions about the relationship between accounts of language learning and the origins of similarities across languages.     

Capacities underlying word learning

Children are strikingly good at learning the meanings of words. Current controversy focuses on the relative importance of different capacities in this learning process including principles of association, low-level attentional mechanisms, special word learning constraints, syntactic cues and theory of mind. We argue that children succeed at word learning because they possess certain conceptual biases about the external world, the ability to infer the referential intentions of others and an appreciation of syntactic cues to word meaning. Support for this view comes from studies exploring the phenomena of fast mapping, the whole object bias, the acquisition of names for entities belonging to different ontological kinds and the effect of lexical contrast. Word learning is not the result of a general associative learning process, nor does it involve specialized constraints. The ability to learn the meanings of words depends on a number of capacities, some of which are specific to language and unique to humans, others of which are potentially shared with other species.     

Using Speakers' Referential Intentions to Model Early Cross-Situational Word Learning

ABSTRACT— Word learning is a "chicken and egg" problem. If a child could understand speakers' utterances, it would be easy to learn the meanings of individual words, and once a child knows what many words mean, it is easy to infer speakers' intended meanings. To the beginning learner, however, both individual word meanings and speakers' intentions are unknown. We describe a computational model of word learning that solves these two inference problems in parallel, rather than relying exclusively on either the inferred meanings of utterances or cross-situational word-meaning associations. We tested our model using annotated corpus data and found that it inferred pairings between words and object concepts with higher precision than comparison models. Moreover, as the result of making probabilistic inferences about speakers' intentions, our model explains a variety of behavioral phenomena described in the word-learning literature. These phenomena include mutual exclusivity, one-trial learning, cross-situational learning, the role of words in object individuation, and the use of inferred intentions to disambiguate reference.     

Weighted constraints in generative linguistics

Harmonic Grammar (HG) and Optimality Theory (OT) are closely related formal frameworks for the study of language. In both, the structure of a given language is determined by the relative strengths of a set of constraints. They differ in how these strengths are represented: as numerical weights (HG) or as ranks (OT). Weighted constraints have advantages for the construction of accounts of language learning and other cognitive processes, partly because they allow for the adaptation of connectionist and statistical models. HG has been little studied in generative linguistics, however, largely due to influential claims that weighted constraints make incorrect predictions about the typology of natural languages, predictions that are not shared by the more popular OT. This paper makes the case that HG is in fact a promising framework for typological research, and reviews and extends the existing arguments for weighted over ranked constraints.     

Frames of reference in spatial language acquisition

Languages differ in how they encode spatial frames of reference. It is unknown how children acquire the particular frame-of-reference terms in their language (e.g., left/right, north/south). The present paper uses a word-learning paradigm to investigate 4-year-old English-speaking children’s acquisition of such terms. In Part I, with five experiments, we contrasted children’s acquisition of novel word pairs meaning left-right and north-south to examine their initial hypotheses and the relative ease of learning the meanings of these terms. Children interpreted ambiguous spatial terms as having environment-based meanings akin to north and south, and they readily learned and generalized north-south meanings. These studies provide the first direct evidence that children invoke geocentric representations in spatial language acquisition. However, the studies leave unanswered how children ultimately acquire “left” and “right.” In Part II, with three more experiments, we investigated why children struggle to master body-based frame-of-reference words. Children successfully learned “left” and “right” when the novel words were systematically introduced on their own bodies and extended these words to novel (intrinsic and relative) uses; however, they had difficulty learning to talk about the left and right sides of a doll. This difficulty was paralleled in identifying the left and right sides of the doll in a non-linguistic memory task. In contrast, children had no difficulties learning to label the front and back sides of a doll. These studies begin to paint a detailed account of the acquisition of spatial terms in English, and provide insights into the origins of diverse spatial reference frames in the world’s languages.     

A review of brain-based neuro-cognitive models

We present a survey of recent brain-based neural network models of various aspects of cognition, covering learning and episodic memory, working memory, attention control, goal formation, language learning and use and consciousness. These models are assessed in the light of their explanation of experimental data, as well as the possible help they have provided in understanding global cognitive processing principles. The paper concludes with a discussion of how the models discussed might help to guide the creation of an artificial cognitive brain.     

Interactions between statistical and semantic information in infant language development

Infants can use statistical regularities to form rudimentary word categories (e.g. noun, verb), and to learn the meanings common to words from those categories. Using an artificial language methodology, we probed the mechanisms by which two types of statistical cues (distributional and phonological regularities) affect word learning. Because linking distributional cues vs. phonological information to semantics make different computational demands on learners, we also tested whether their use is related to language proficiency. We found that 22-month-old infants with smaller vocabularies generalized using phonological cues; however, infants with larger vocabularies showed the opposite pattern of results, generalizing based on distributional cues. These findings suggest that both phonological and distributional cues marking word categories promote early word learning. Moreover, while correlations between these cues are important to forming word categories, we found infants' weighting of these cues in subsequent word-learning tasks changes over the course of early language development.     

Maturational Constraints on Language Learning

This paper suggests that there are constraints on learning required to explain the acquisition of language, in particular, maturational constraints. First, empirical evidence for this claim is reviewed. The evidence from several studies of both first and second language acquisition suggests that normal language learning occurs only when exposure to the language begins early in life. With exposure beginning later in life, asymptotic performance in the language declines: the effects over age of first exposure are approximately linear through childhood, with a flattening of the function in adulthood. These outcomes argue that some type of constraints ensuring successful language learning exist early in life, and weaken with increasing maturation. Second, two hypotheses are considered as to the nature of these maturational changes. One hypothesis is that constraints on learning particular to language acquisition undergo maturational decay. A second hypothesis, which is considered in more detail, suggests that language learning abilities decline because of the expansion of nonlinguistic cognitive abilities.     

The bilingual advantage in novel word learning

The present study examined whether bilingualism facilitates acquisition of novel words in adults with different language histories. Word-learning performance was tested in monolingual English speakers, early English—Spanish bilinguals, and early English—Mandarin bilinguals. Novel words were phonologically unfamiliar to all participants, and they were acquired in association with their English translations. At testing, both bilingual groups outperformed the monolingual group. These findings indicate that bilingualism facilitates word-learning performance in adults, and they suggest a general bilingual advantage for novel word learning.     

Sensitivity to sampling in Bayesian word learning

We report a new study testing our proposal that word learning may be best explained as an approximate form of Bayesian inference (Xu & Tenenbaum, in press). Children are capable of learning word meanings across a wide range of communicative contexts. In different contexts, learners may encounter different sampling processes generating the examples of word-object pairings they observe. An ideal Bayesian word learner could take into account these differences in the sampling process and adjust his/her inferences about word meaning accordingly. We tested how children and adults learned words for novel object kinds in two sampling contexts, in which the objects to be labeled were sampled either by a knowledgeable teacher or by the learners themselves. Both adults and children generalized more conservatively in the former context; that is, they restricted the label to just those objects most similar to the labeled examples when the exemplars were chosen by a knowledgeable teacher, but not when chosen by the learners themselves. We discuss how this result follows naturally from a Bayesian analysis, but not from other statistical approaches such as associative word-learning models.     

注意和记忆对孤独症儿童词汇发展滞后的影响

孤独症语言障碍的表现之一是词汇发展滞后, 可能与其注意和记忆损伤有关。当前研究结果表明, 孤独症儿童在学习词汇时难以利用社会注意提供的有效信息, 且其注意易受到无关刺激干扰, 这可能导致其形成的物体-词汇的联结不稳定, 影响其进一步将这种联结整合到心理词典并保存在记忆中。未来研究应探究联合注意影响孤独症儿童词汇学习的发展轨迹和机制, 儿童的词汇知识对其词汇记忆的影响, 并关注自然场景中孤独症儿童的词汇学习过程和个体差异。     

Word learning in dogs?

In a recent paper, Kaminski, Call and Fischer report pioneering research on word-learning in a dog. In this commentary we suggest ways of distinguishing referential word use from mere association. We question whether the dog is reasoning by exclusion and, if so, compare three explanations - learned heuristics, default assumptions, and pragmatic reasoning - as they apply to children and might apply to dogs. Kaminski et al.'s work clearly raises important questions about the origins and basis of word learning and social cognition.     

Reflections on how color term acquisition is constrained

Compared with object word learning, young children typically find learning color terms to be a difficult linguistic task. In this reflections article, I consider two questions that are fundamental to investigations into the developmental acquisition of color terms. First, I consider what constrains color term acquisition and how stable these constraints are over culture and age. I review recent studies that have identified conceptual, attentional, and linguistic constraints that seemingly operate on the acquisition of color terms compared with object words and the differential acquisition of color terms within a given language. Second, I consider whether these constraints are specific to the acquisition of color terms or whether they reflect more general constraints that operate on other classes of lexical terms such as different abstract property terms.     

Evading conceptual self-annihilation: Some implications of Albert Bandura's theory of the self-system for the status of psychology

There are two cognitive revolutions occuring in psychology. The first in the fields of perception, memory, language and thought uses information processing and computer models of cognition. The second in learning theory and behavioral therapy uses folk psychological concepts of cognition. To test the philosophical compatibility of these cognitive revolutions I examine the implications of Albert Bandura's theory of the self-system for the disputed question of the status of folk psychology. I argue that if the theory of the self-system is correct, then strong eliminative materialist claims about the falsity of folk psychology are disconfirmed, functionalist homunculi hypotheses need modification, and certain constraints are imposed on neurophysiological accounts of the self. Thus I conclude that the cognitive revolution in learning theory and behavior therapy is making its own distinctive contribution to the new philosophical psychology emerging from the cognitive revolution in psychology as a whole.     

Cross-modal working memory binding and L1-L2 word learning

The ability to create temporary binding representations of information from different sources in working memory has recently been found to relate to the development of monolingual word recognition in children. The current study explored this possible relationship in an adult word-learning context. We assessed whether the relationship between cross-modal working memory binding and lexical development would be observed in the learning of associations between unfamiliar spoken words and their semantic referents, and whether it would vary across experimental conditions in first- and second-language word learning. A group of English monolinguals were recruited to learn 24 spoken disyllable Mandarin Chinese words in association with either familiar or novel objects as semantic referents. They also took a working memory task in which their ability to temporarily bind auditory-verbal and visual information was measured. Participants’ performance on this task was uniquely linked to their learning and retention of words for both novel objects and for familiar objects. This suggests that, at least for spoken language, cross-modal working memory binding might play a similar role in second language-like (i.e., learning new words for familiar objects) and in more native-like situations (i.e., learning new words for novel objects). Our findings provide new evidence for the role of cross-modal working memory binding in L1 word learning and further indicate that early stages of picture-based word learning in L2 might rely on similar cognitive processes as in L1.