Monster character profiling and Chinese classifier cognition: linguistic and conceptual perspectives

In this research, we explored the construal of mythical creatures created out of ordinary animals by approaching the selection of Chinese classifiers zhi and ge. Monster figures as such are viewed ambiguously as an animal or as a novel entity. From the lexico-semantic perspective, we surveyed the co-occurrences of these two classifiers with two Chinese labels commonly used for monsters: guaishou and guaiwu, which provide lexical cues for the finely nuanced difference in (non-)/animate readings respectively. From the conceptual perspective, we examined the acceptability as to the uses of these two classifiers for the monster figures, the structures of which incurred different effects of novelty. We found that the (non-)/animal-like interpretation is encoded by zhi and ge at both the linguistic and conceptual layers. The findings of our work have implications for the relation of language and thought, our understanding of fictional world entities, and our understanding of classifier cognition.     

Heuristics in learning classifiers: The acquisition of the classifier system and its implications for the nature of lexical acquisition

Classifiers are like nouns in that they classify entities in the world into lexical categories. However, the lexical nature of the classifier system is very different from that of nouns. We discuss how Japanese and Chinese children learn the meanings of classifiers. We focus on two specific questions: How classifier acquisition is different from noun acquisition; and what the prerequisites are for spontaneously extracting the meanings of classifiers. It is shown that children are very conservative in assigning meaning to classifiers. The pace of learning largely depends on semantic complexity, across languages and within each language. Furthermore, we suspect that learning the meanings of classifiers requires a certain cognitive ability – an ability to synthesize pieces of partial knowledge and form them into a cohesive whole. It may be only when children have developed such an ability that they are able to extract the complex semantic rules of classifiers on their own. We conclude that children take very different routes in learning nouns and classifiers: Unlike noun acquisition, classifier acquisition is guided by a slow, bottom-up process.     

Nested diagnostic classification models for multiple-choice items

This study proposes and evaluates a diagnostic classification model framework for multiple-choice items. Models in the proposed framework have a two-level nested structure which allows for binary scoring (for correctness) and polytomous scoring (for distractors) at the same time. One advantage of these models is that they can provide distractor information while maintaining the statistical properties of the correct response option. We evaluated parameter recovery through a simulation study using Hamiltonian Monte Carlo algorithms in Stan. We also discussed three approaches to implementing the proposed modelling framework for different purposes and testing scenarios. We illustrated those approaches and compared them with a binary model and a traditional nominal model through an operational study.     

Psycholinguistic research in language intervention programming: The pronoun system

This paper presents a linguistic analysis of the pronoun system in terms of binary semantic and syntactic features. A model is presented which indicates how these features serve to differentiate among the members of this set. A hypothetical order of development of these features is proposed, supported by developmental data. Suggestions are made regarding the possible relationship of the pronoun system to other areas of language. Finally, proposals are presented for the development of pronoun training procedures within the context of language training programs.This research was supported in part by Speech and Hearing grant HD 05088 and in part by CORE grant HD 02528.     

Nearest vertex attraction for actively reducing loss

In recent years the neural networks have received considerable success on classification tasks. The randomly initialized network is generally expected to adjust the weights for reducing the loss, which is also indicated as the total distance from the samples to the corresponding vertices. However, the output space of the conventional neural networks suffers from the fixed relation between the labels and vertices during learning. This case forces the mapped points around the unexpected vertex across the decision boundary into the neighborhood of the correct vertex, and simultaneously the boundary points cannot obtain the substantial rectification. Therefore, this study proposes a novel nearest vertex attraction (NVA) to actively adjust the relation between the categories and the output vertices for improving the neural network classifiers. The best relation allows that the data points can be attracted by the nearest vertices to minimize the total moving distances. In this way, the mapped points that are near to the vertices and the decision boundaries obtain the decent management. We evaluated the NVA with several conventional classification techniques and other neural network classifiers on 12 public UCI datasets. The numerical experiments demonstrate that the proposed method improves performance of the neural network classifiers on the involved benchmarks.     

Information about the logical structure of a category affects generalization

This paper considers whether information about the logical structure of a category affects how people generalize. We carried out three experiments with the following structure: participants were first presented with a set of training items, and were subsequently asked to decide whether new items belonged to the same category as the training items. Each experiment had two conditions that differed only in terms of the category label provided for the training items; different category labels conveyed different information about the logical structure of the category to which the training items were supposed to belong. In all cases, participants' generalization was greatly affected by such information. Our results suggest that people make the default assumption that category labels correspond to groupings of highly similar objects.     

Linguistic influences on visual memory

The Whorf-Sapir hypothesis has raised considerable controversy in the literatures of psychology and anthropology. Several misconceptions of the hypothesis are reviewed, and the hypothesis was experimentally supported in a visual reproduction paradigm. Subjects were first given label training for a set of figures, and were then asked to recall by drawing the shapes. Training with categorized labels resulted in a 25% improvement in recall when compared to a condition with nonword (paralog) labels. Even stronger evidence of linguistic influence on visual memory was obtained by examining the order of recall. The conceptual relationships among labels strongly influenced the sequence of reproductions.     

A New Multilevel CART Algorithm for Multilevel Data with Binary Outcomes

The multilevel logistic regression model (M-logit) is the standard model for modeling multilevel data with binary outcomes. However, many assumptions and restrictions should be considered when applying this model for unbiased estimation. To overcome these limitations, we proposed a multilevel CART (M-CART) algorithm which combines the M-logit and single level CART (S-CART) within the framework of the expectation-maximization. Simulation results showed that the proposed M-CART provided substantial improvements on classification accuracy, sensitivity, and specific over the M-logit, S-CART, and single level logistic regression model when modeling multilevel data with binary outcomes. This benefit of using M-CART was consistently found across different conditions of sample size, intra-class correlation, and when relationships between predictors and outcomes were nonlinear and nonadditive.     

Efficiency investigation from shallow to deep neural network techniques in human activity recognition

In the last years, several researchers measured different recognition rates with different artificial neural network (ANN) techniques on public data sets in the human activity recognition (HAR) problem. However an overall investigation does not exist in the literature and the efficiency of complex and deeper ANNs over shallow networks is not clear. The purpose of this paper is to investigate the recognition rate and time requirement of different kinds of ANN approaches in HAR. This work examines the performance of shallow ANN architectures with different hyper-parameters, ANN ensembles, binary ANN classifier groups, and convolutional neural networks on two public databases. Although the popularity of binary classifiers, classifier ensembles and deep learning have been significantly increasing, this study shows that shallow ANNs with appropriate hyper-parameters in combination with extracted features can reach similar or higher recognition rate in less time than other artificial neural network methods in HAR. With a well-tuned ANN we outperformed all previous results on two public databases. Consequently, instead of the more complex ANN techniques, the usage of simple ANN with two or three layers can be an appropriate choice for activity recognition.     

A performance comparison between shallow and deeper neural networks supervised classification of tomosynthesis breast lesions images

Computer Aided Decision (CAD) systems, based on 3D tomosynthesis imaging, could support radiologists in classifying different kinds of breast lesions and then improve the diagnosis of breast cancer (BC) with a lower X-ray dose than in Computer Tomography (CT) systems.In previous work, several Convolutional Neural Network (CNN) architectures were evaluated to discriminate four different classes of lesions considering high-resolution images automatically segmented: (a) irregular opacity lesions, (b) regular opacity lesions, (c) stellar opacity lesions and (d) no-lesions. In this paper, instead, we use the same previously extracted relevant Regions of Interest (ROIs) containing the lesions, but we propose and evaluate two different approaches to better discriminate among the four classes.In this work, we evaluate and compare the performance of two different frameworks both considering supervised classifiers topologies. The first framework is feature-based, and consider morphological and textural hand-crafted features, extracted from each ROI, as input to optimised Artificial Neural Network (ANN) classifiers. The second framework, instead, considers non-neural classifiers based on automatically computed features evaluating the classification performance extracting several sets of features using different Convolutional Neural Network models.Final results show that the second framework, based on features computed automatically by CNN architectures performs better than the first approach, in terms of accuracy, specificity, and sensitivity.     

Vocal learning in the Grey parrot (Psittacus erithacus): effects of species identity and number of trainers

Grey parrots (Psittacus erithacus) learn referential English labels when they view and interact with 2 humans who model vocal labeling and who demonstrate referentiality and functionality of a label (I.M. Pepperberg, 1990a). To test if both trainers are necessary, the authors contrasted 2-trainer modeling with training by 1 human who presented targeted labels to a bird in concert with appropriate items, who asked questions, and who would reward attempts at the label with the item. The bird was also tutored by either 1 or 2 interactive humans in conjunction with a conspecific who already used referential labels. Referential labels were learned from multiple live tutors but not a single trainer. Presence of a conspecific enhanced learning compared with single-trainer sessions but did not affect acquisition in 2-human sessions. Specific aspects of paired tutoring seem critical for acquiring referential vocal labels.     

Questioning and Creating: A Model for Curriculum Reform

ABSTRACT Creativity research has unsettled education by newly defining neglected abilities, and by proposing new methods like training in questioning abilities, but conventional methods do not encourage students to learn or think by asking good questions. We need to cultivate the full spectrum of talents by bridging the research—classroom practice gap. A proposed three-dimensional model for analyzing curricula, courses, and classes would show us where we are not doing a good job in teaching students to think while we are simultaneously teaching them subject-matter content. Content is best learned as a by-product of thinking or learning processes. The proposed model capitalizes on transfer effects of both processes and content, and may be of value in working, not only with those highest in a talent, but also with the educationally deprived. Every student should learn to use each of his abilities at some point in the educational process, but different teachers may specialize in various contents, processes, or methods.     

Training sequence effects on emergent conditional discriminations: Replication and extension to selection-based training

We examined the replicability and generality of a previously reported training sequence effect on emergent conditional discriminations in the intraverbal naming task. In Experiment 1, a tact–intraverbal (TI) group learned first to vocally label 6 visual patterns and then to intraverbally relate pairs of verbal labels, whereas an intraverbal–tact (IT) group received the same training in the opposite sequence. Emergent conditional discriminations among pattern stimuli were assessed in match-to-sample (MTS) format. Experiment 2 was identical, except vocal tact and intraverbal training were replaced with selection-based training in which the verbal labels were text stimuli. Compared to the IT sequence, the TI sequence resulted in greater mean accuracy at test (Experiment 1), higher yields (Experiment 2), and shorter reaction times (Experiment 2). Experiment 2 data suggested the TI group's performance might be less dependent on intact intraverbal relations relative to the IT group, but related to participants' reports of visualization during intraverbal training. The results suggest the sequence effect is replicable and occurs in experimental preparations commonly used to study derived stimulus relations. They also provide novel support for the hypothesis that participant behavior during training alters sources of stimulus control available at test.     

Avoid violence,rioting, and outrage; approach celebration,delight, and strength: Using large text corpora to compute valence,arousal, and the basic emotions

Ever since Aristotle discussed the issue in Book II of his Rhetoric, humans have attempted to identify a set of “basic emotion labels”. In this paper we propose an algorithmic method for evaluating sets of basic emotion labels that relies upon computed co-occurrence distances between words in a 12.7-billion-word corpus of unselected text from USENET discussion groups. Our method uses the relationship between human arousal and valence ratings collected for a large list of words, and the co-occurrence similarity between each word and emotion labels. We assess how well the words in each of 12 emotion label sets—proposed by various researchers over the past 118 years—predict the arousal and valence ratings on a test and validation dataset, each consisting of over 5970 items. We also assess how well these emotion labels predict lexical decision residuals (LDRTs), after co-varying out the effects attributable to basic lexical predictors. We then demonstrate a generalization of our method to determine the most predictive “basic” emotion labels from among all of the putative models of basic emotion that we considered. As well as contributing empirical data towards the development of a more rigorous definition of basic emotions, our method makes it possible to derive principled computational estimates of emotionality—specifically, of arousal and valence—for all words in the language.     

REPLACING MALADAPTIVE SPEECH WITH VERBAL LABELING RESPONSES: AN ANALYSIS OF GENERALIZED RESPONDING

We taught three mentally handicapped students to answer questions with verbal labels and evaluated the generalized effects of this training on their maladaptive speech (e.g., echolalia) and correct responding to untrained questions. The students received cues-pause-point training on an initial question set followed by generalization assessments on a different set in another setting. Probes were conducted on novel questions in three other settings to determine the strength and spread of the generalization effect. A multiple baseline across subjects design revealed that maladaptive speech was replaced with correct labels (answers) to questions in the training and all generalization settings. These results replicate and extend previous research that suggested that cues-pause-point procedures may be useful in replacing maladaptive speech patterns by teaching students to use their verbal labeling repertoires.     

Training and testing music skills in a boy with autism using a matching‐to‐sample format

A 16‐year old boy with autism was taught music skills using a matching to sample procedure. He was trained and subsequently tested for the formation of four 4‐member classes, including different visual music stimuli, and Norwegian and Vietnamese labels for different major and minor chords. Four different stimuli sets were trained both in one‐to‐many (OTM) and many‐to‐one (MTO) training structures. Further, we explored if the reaction times to comparison stimuli increased from training to testing. Results showed that the participant formed equivalence classes with music relations. Furthermore, there were small differences only between OTM and MTO with respect to stimulus equivalence responding. The reaction times to comparison stimuli increased from training to testing, and were most pronounced for the equivalence trials. Copyright © 2010 John Wiley & Sons, Ltd.     

The effects of attention-directing training on recognition memory task performance of three-year-old children

Thirty-eight 3-year-old children served as subjects in an investigation of recognition memory in which schematic faces differing only in the orientation of the eyes were employed as stimuli. A pretest was administered to all children, after which the two experimental groups received training in either attention to the distinctive feature of the training stimuli (also schematic faces) by means of a matching task, or in labeling the faces according to how they looked (sleepy, happy, sad, mad) and in using the labels to perform a matching task. After the training session all children were given a posttest on recognition memory of the faces. The verbally trained group obtained significantly higher scores on the posttest than either the feature or control groups. These results indicate that although the children were able to discriminate the faces, evidenced in their ease of performance on the training tasks, they were not able to use this knowledge unless given training in attaching labels to the stimuli, which enabled them to store the information for later use. Results are discussed in light of Gibson's (1969) theory of the developmental interrelations of cognitive processes.     

Preschoolers trust particular informants when learning new names and new morphological forms

Across three studies, we investigated whether 4‐year‐olds would trust a previously reliable informant when learning novel morphological forms. In Experiment 1, children (N= 16) were presented with two informants: one who correctly named familiar objects and another who named them incorrectly. Children were invited to turn to these informants when learning novel labels and morphological forms. The majority of children chose the previously correct labeller when learning novel label and morphology. In Experiment 2, children (N= 16) were presented with an informant who used familiar plurals correctly and one who used them incorrectly. Children chose the previously correct morphologist when learning novel labels and past tense forms. Thus, children track both semantic and morphological accuracy. In Experiment 3, some children (N= 16) were presented with two informants who differed in naming accuracy, whereas others (N= 16) were presented with two informants who differed in morphological accuracy. To forestall any risk of experimenter cuing, one experimenter blind to the training children had received, tested children with novel labels and morphology. The results replicated those of Experiments 1 and 2. Implications for how children's trust in an informant might play a role in their acquisition of morphological forms are discussed.     

Associative factors in free recall

An experiment on the free recall of 18-word lists is reported in which variations were made in associative frequency, word frequency and the number, and therefore size, of the presented groups. Recall efficiency was positively related to increasing associative frequency and larger presented group size. Measures of clustering were obtained along with subjects' own reports on the associative labels helpful to them in organizing their recall. A two-level storage model is proposed to account for the results.     

Comparison of sign versus verbal symbol training in retarded adults

Sign language as an alternative or as an augmentive system to verbal language training in the mentally retarded is in widespread use. This study began an exploration of the relationship between sign and verbal learning in 10 institutionalized severely mentally retarded adults. Three experimental groups were taught color labels. Three persons received sign training only, 4 more received verbal training only, and last 3 received combined verbal and sign training. Sign labels tended to be learned more efficiently than verbal labels by this small group. Combined sign and verbal training improved verbal learning whereas sign learning was not improved through the combined approach. Replication and extension of this preliminary work with a larger and more representative sample is needed.