Shohei Hidaka

Body Parts and Early-Learned Verbs.

This article reports the structure of associations among 101 common verbs and body parts. The verbs are those typically learned by children learning English prior to 3 years of age. In a free association task, 50 adults were asked to provide the single body part that came to mind when they thought of each verb. Analyses reveal highly systematic and structured patterns of associations that are also related to the normative age of acquisition of the verbs showing a progression from verbs associated with actions by the mouth, to verbs strongly associated with actions by hand and arm, to verbs not so strongly associated with any one body part. The results have implications for proposals about embodied verb meaning and also for processes of early verb learning.

Sound Symbolism Facilitates Word Learning in 14-Month-Olds

Sound symbolism, or the nonarbitrary link between linguistic sound and meaning, has often been discussed in connection with language evolution, where the oral imitation of external events links phonetic forms with their referents (e.g., Ramachandran & Hubbard, 2001). In this research, we explore whether sound symbolism may also facilitate synchronic language learning in human infants. Sound symbolism may be a useful cue particularly at the earliest developmental stages of word learning, because it potentially provides a way of bootstrapping word meaning from perceptual information. Using an associative word learning paradigm, we demonstrated that 14-month-old infants could detect Köhler-type (1947) shape-sound symbolism, and could use this sensitivity in their effort to establish a word-referent association.

A Single Word in a Population of Words

Carey and Bartlett introduced a new method for studying lexical development, one of presenting the child with a word and a single context of use and asking what was learned from that one encounter. They also reported a then new finding; that is, by using what they already knew about previously learned words, young children could narrow the range of possibilities for likely meanings in a single encounter. This papers honors that original contribution and the robust literature and set of phenomena it generated by considering how newly learned categories must fit into a population of already learned categories. This paper presents an overview of Packing Theory, a formal geometrical analysis of how local interactions in a large population of categories create a global structure of feature relevance such that near categories in the population of have similar generalization patterns. The implications of these ideas for learning from a single encounter, their relation to the evidence of artificial word learning studies, and new predictions are discussed.

A data-driven paradigm to understand multimodal communication in human-human and human-robot interaction

Data-driven knowledge discovery is becoming a new trend in various scientific fields. In light of this, the goal of the present paper is to introduce a novel framework to study one interesting topic in cognitive and behavioral studies – multimodal communication between human-human and human-robot interaction. We present an overall solution from data capture, through data coding and validation, to data analysis and visualization. In data collection, we have developed a multimodal sensing system to gather fine-grained video, audio and human body movement data. In data analysis, we propose a hybrid solution based on visual data mining and information-theoretic measures. We suggest that this data-driven paradigm will lead not only to breakthroughs in understanding multimodal communication, but will also serve as a successful case study to demonstrate the promise of data-intensive discovery which can be applied in various research topics in cognitive and behavioral studies.

Quantitative linking hypotheses for infant eye movements.

The study of cognitive development hinges, largely, on the analysis of infant looking. But analyses of eye gaze data require the adoption of linking hypotheses: assumptions about the relationship between observed eye movements and underlying cognitive processes. We develop a general framework for constructing, testing, and comparing these hypotheses, and thus for producing new insights into early cognitive development. We first introduce the general framework – applicable to any infant gaze experiment – and then demonstrate its utility by analyzing data from a set of experiments investigating the role of attentional cues in infant learning. The new analysis uncovers significantly more structure in these data, finding evidence of learning that was not found in standard analyses and showing an unexpected relationship between cue use and learning rate. Finally, we discuss general implications for the construction and testing of quantitative linking hypotheses. MATLAB code for sample linking hypotheses can be found on the first authors website.

Packing: A geometric analysis of feature selection and category formation

This paper presents a geometrical analysis of how local interactions in a large population of categories packed into a feature space create a global structure of feature relevance. The theory is a formal proof that the joint optimization of discrimination and inclusion creates a smooth space of categories such that near categories in the similarity space have similar generalization gradients. Packing theory offers a unified account of several phenomena in human categorization including the differential importance of different features for different kinds of categories, the dissociation between judgments of similarity and judgments of category membership, and childrens ability to generalize a category from very few examples.

Spatio-Temporal Symbolization of Multidimensional Time Series

The present study proposes a new symbolization algorithm for multidimensional time series. We view temporal sequences as observed data generated by a dynamical system, and therefore the goal of symbolization is to estimate symbolic sequences that minimize loss of information, which is called generating partition in nonlinear physics. In order to utilize the theoretical property of symbol dynamics in data mining, our algorithm estimates symbols on multivariate time series by integrating both spatial and temporal information and selecting those dimensions in multidimensional time series containing useful information. Probabilistic symbolic sequences derived from our symbolization method can be used in various supervised and unsupervised data-mining tasks. To demonstrate this, the algorithm is evaluated by applying it to both simulated data and a real-world dataset. In both cases, the new algorithm outperforms its alternative approaches.

A computational model associating learning process, word attributes, and age of acquisition.

We propose a new model-based approach linking word learning to the age of acquisition (AoA) of words; a new computational tool for understanding the relationships among word learning processes, psychological attributes, and word AoAs as measures of vocabulary growth. The computational model developed describes the distinct statistical relationships between three theoretical factors underpinning word learning and AoA distributions. Simply put, this model formulates how different learning processes, characterized by change in learning rate over time and/or by the number of exposures required to acquire a word, likely result in different AoA distributions depending on word type. We tested the model in three respects. The first analysis showed that the proposed model accounts for empirical AoA distributions better than a standard alternative. The second analysis demonstrated that the estimated learning parameters well predicted the psychological attributes, such as frequency and imageability, of words. The third analysis illustrated that the developmental trend predicted by our estimated learning parameters was consistent with relevant findings in the developmental literature on word learning in children. We further discuss the theoretical implications of our model-based approach.

Analyzing multimodal time series as dynamical systems

We propose a novel approach to discovering latent structures from multimodal time series. We view a time series as observed data from an underlying dynamical system. In this way, analyzing multimodal time series can be viewed as finding latent structures from dynamical systems. In light this, our approach is based on the concept of generating partition which is the theoretically best symbolization of time series maximizing the information of the underlying original continuous dynamical system. However, generating partition is difficult to achieve for time series without explicit dynamical equations. Different from most previous approaches that attempt to approximate generating partition through various deterministic symbolization processes, our algorithm maintains and estimates a probabilistic distribution over a symbol set for each data point in a time series. To do so, we develop a Bayesian framework for probabilistic symbolization and demonstrate that the approach can be successfully applied to both simulated data and empirical data from multimodal agent-agent interactions. We suggest this unsupervised learning algorithm has a potential to be used in various multimodal datasets as first steps to identify underlying structures between temporal variables.

A connectionist account of ontological boundary shifting

Previous research on children’s categorizations has suggested that children use perceptual and conceptual knowledge to generalize object names. In particular, the relation between ontological categories and linguistic categories appears to be a critical cue to learning object categories. However, the mechanism underlying this relation remains unclear. Here we propose a connectionist model for the acquisition of ontological knowledge by learning linguistic categories of entities. The results suggest that linguistic cues help children attend to specific perceptual properties.