汉字形声字识别中义符和声符的家族效应

采用词汇判断任务和ERP技术, 操纵义符和声符的家族大小, 考察形声字识别中义符和声符的家族效应。结果发现：(1)当义符为大家族时, 大家族声符字比小家族声符字诱发了更小的P200波幅; 当义符为小家族时, 大家族声符字和小家族声符字诱发的P200波幅差异不显著。(2)大家族声符字比小家族声符字诱发了更大的N400波幅, 但当义符为大家族时, 大、小家族声符字之间的差异比义符为小家族时更加显著。这说明, 义符和声符的家族大小共同影响形声字的识别, 声符的家族效应受义符的家族大小调节。研究表明, 在词汇识别的不同阶段, 义符和声符对资源的获取处于变化状态, 这取决于两者的呈现率差异。整个研究表明, 在形声字的识别中, 义符与声符的作用存在着博弈, 二者作用的此消彼长与加工阶段和家族大小有关。     

Semantic segmentation via highly fused convolutional network with multiple soft cost functions

Semantic image segmentation is one of the most challenging tasks in computer vision. In this paper, we propose a highly fused convolutional network, which consists of three parts: downsampling, fused upsampling and multiple predictions. We adopt a VGG-net based downsampling structure, followed by multiple steps of upsampling. Feature maps in each pair of corresponding pooling layers and unpooling layers are combined. We also bring out multiple pre-outputs, each is generated from an unpooling layer by a one-step upsampling operation. Finally, we concatenate these pre-outputs to get the final output. As a result, our proposed network makes high use of the feature information by fusing and reusing features in low layers. In addition, when training our model, we add multiple soft cost functions on pre-outputs and the final output. In this way, we can reduce the loss reduction in backpropagation. We evaluate our model on three public segmentation datasets: CamVid, PASCAL VOC, and ADE20K. We achieve considerable segmentation performance on PASCAL VOC dataset and ADE20K dataset. Especially on CamVid dataset, we achieve state-of-the-art performance.     

The semantic pointer theory of emotion: Integrating physiology,appraisal, and construction

Emotion theory needs to explain the relationship of language and emotions, and the embodiment of emotions, by specifying the computational mechanisms underlying emotion generation in the brain. We used Chris Eliasmith’s Semantic Pointer Architecture to develop POEM, a computational model that explains numerous important phenomena concerning emotions, including how some stimuli generate immediate emotional reactions, how some emotional reactions depend on cognitive evaluations, how bodily states influence the generation of emotions, how some emotions depend on interactions between physiological inputs and cognitive appraisals, and how some emotional reactions concern syntactically complex representations. We contrast our theory with current alternatives, and discuss some possible applications to individual and social emotions.     

Musical structure modulates semantic priming in vocal music

It has been shown that harmonic structure may influence the processing of phonemes whatever the extent of participants' musical expertise [Bigand, E., Tillmann, B., Poulin, B., D'Adamo, D. A., & Madurell, F. (2001). The effect of harmonic context on phoneme monitoring in vocal music. Cognition, 81, B11-B20]. The present study goes a step further by investigating how musical harmony may potentially interfere with the processing of words in vocal music. Eight-chord sung sentences were presented, their last word being either semantically related (La girafe a un tres grand cou, The giraffe has a very long neck) or unrelated to the previous linguistic context (La girafe a un tres grand pied, The giraffe has a very long foot). The target word was sung on a chord that acted either as a referential tonic chord or as a congruent but less referential subdominant chord. Participants performed a lexical decision task on the target word. A significant interaction was observed between semantic and harmonic relatedness suggesting that music modulates semantic priming in vocal music. Following Jones' dynamic attention theory, we argue that music can modulate semantic priming in vocal music, by modifying the allocation of attentional resource necessary for linguistic computation.     

Evidence for dissociable neural mechanisms underlying inference generation in familiar and less-familiar scenarios

In this study, we investigated whether the left and right hemispheres are differentially involved in causal inference generation. Participants read short inference-promoting texts that described either familiar or less-familiar scenarios. After each text, they performed a lexical decision on a letter string (which sometimes constituted an inference-related word) presented directly to the left or right hemisphere. Response-time results indicated that hemisphere of direct presentation interacted with type of inference scenario. When test stimuli were presented directly to the left hemisphere, lexical decisions were facilitated following familiar but not following less-familiar inference scenarios, whereas when test stimuli were presented directly to the right hemisphere, facilitation was observed in both familiar and less-familiar conditions. Thus, inferences may be generated in different ways depending on which of two dissociable neural subsystems underlies the activation of background information.     

Hemispheric processing of anaphoric inferences: the activation of multiple antecedents

This research investigates the hemispheric processing of anaphors when readers activate multiple antecedents. Participants read texts promoting an anaphoric inference and performed a lexical decision task to inference-related target words that were consistent (Experiment 1) or inconsistent (Experiment 2) with the text. These targets were preceded by constrained or less constraining text and were presented to participants' right visual field-left hemisphere or to their left visual field-right hemisphere. In Experiment 1, both hemispheres showed facilitation for consistent antecedents and the right hemisphere showed an advantage over the left hemisphere in processing antecedents when preceded by less constrained text. In Experiment 2, the left hemisphere only showed negative facilitation for inconsistent antecedents. When readers comprehend text with multiple antecedents: both hemispheres process consistent information, the left hemisphere inhibits inconsistent information, and the right hemisphere processes less constrained information.     

A memory-based theory of verbal cognition

The syntagmatic paradigmatic model is a distributed, memory-based account of verbal processing. Built on a Bayesian interpretation of string edit theory, it characterizes the control of verbal cognition as the retrieval of sets of syntagmatic and paradigmatic constraints from sequential and relational long-term memory and the resolution of these constraints in working memory. Lexical information is extracted directly from text using a version of the expectation maximization algorithm. In this article, the model is described and then illustrated on a number of phenomena, including sentence processing, semantic categorization and rating, short-term serial recall, and analogical and logical inference. Subsequently, the model is used to answer questions about a corpus of tennis news articles taken from the Internet. The model's success demonstrates that it is possible to extract propositional information from naturally occurring text without employing a grammar, defining a set of heuristics, or specifying a priori a set of semantic roles.     

Memory systems of the brain: a brief history and current perspective

The idea that memory is composed of distinct systems has a long history but became a topic of experimental inquiry only after the middle of the 20th century. Beginning about 1980, evidence from normal subjects, amnesic patients, and experimental animals converged on the view that a fundamental distinction could be drawn between a kind of memory that is accessible to conscious recollection and another kind that is not. Subsequent work shifted thinking beyond dichotomies to a view, grounded in biology, that memory is composed of multiple separate systems supported, for example, by the hippocampus and related structures, the amygdala, the neostriatum, and the cerebellum. This article traces the development of these ideas and provides a current perspective on how these brain systems operate to support behavior.     

Mental representations of social status

How do people think about social status? We investigated the nature of social status and number representations using a semantic distance latency test. In Study 1, 21 college students compared words connoting different social status as well as numbers, which served as a control task. Participants were faster at comparing occupations and numbers that were semantically farther apart relative to those more closely related. In Study 2, we examined the semantic distance effect for a social status category, for which participants have as much knowledge of, as with numbers. We asked 15 US Navy Midshipmen to compare the social status associated with different ranks in the Navy as well as compare number magnitudes. Participants were fastest when comparing ranks far in status relative to ranks close in status. These findings reveal that humans have mental representations of social status that share properties with that of number.