Contextual self-organizing map: software for constructing semantic representations

In this article, we introduce a software package that applies a corpus-based algorithm to derive semantic representations of words. The algorithm relies on analyses of contextual information extracted from a text corpus—specifically, analyses of word co-occurrences in a large-scale electronic database of text. Here, a target word is represented as the combination of the average of all words preceding the target and all words following it in a text corpus. The semantic representation of the target words can be further processed by a self-organizing map (SOM; Kohonen, Self-organizing maps, 2001), an unsupervised neural network model that provides efficient data extraction and representation. Due to its topography-preserving features, the SOM projects the statistical structure of the context onto a 2-D space, such that words with similar meanings cluster together, forming groups that correspond to lexically meaningful categories. Such a representation system has its applications in a variety of contexts, including computational modeling of language acquisition and processing. In this report, we present specific examples from two languages (English and Chinese) to demonstrate how the method is applied to extract the semantic representations of words.     

Effects of Emotional Experience for Abstract Words in the Stroop Task

In this study, we examined the effects of emotional experience, a relatively new dimension of emotional knowledge that gauges the ease with which words evoke emotional experience, on abstract word processing in the Stroop task. In order to test the context‐dependency of these effects, we accentuated the saliency of this dimension in Experiment 1A by blocking the stimuli such that one block consisted of the stimuli with the highest emotional experience ratings and the other block consisted of the stimuli with the lowest emotional experience ratings. We attenuated the saliency of this dimension in Experiment 1B by intermixing the stimuli. We observed slower color naming performance for words with higher emotional experience ratings only in Experiment 1A, suggesting that the dimension of emotional experience is an aspect of semantic representation for abstract words but that its influence can be modulated by context. We interpret these results more generally using Vigliocco, Meteyard, Andrews, and Kousta's (2009) framework of semantic representation, and more specifically using Cohen, Dunbar, and McClelland's (1990) model of Stroop task performance.     

Grounding co-occurrence: Identifying features in a lexical co-occurrence model of semantic memory

Lexical co-occurrence models of semantic memory represent word meaning by vectors in a high-dimensional space. These vectors are derived from word usage, as found in a large corpus of written text. Typically, these models are fully automated, an advantage over models that represent semantics that are based on human judgments (e.g., feature-based models). A common criticism of co-occurrence models is that the representations are not grounded: Concepts exist only relative to each other in the space produced by the model. It has been claimed that feature-based models offer an advantage in this regard. In this article, we take a step toward grounding a cooccurrence model. A feed-forward neural network is trained using back propagation to provide a mapping from co-occurrence vectors to feature norms collected from subjects. We show that this network is able to retrieve the features of a concept from its co-occurrence vector with high accuracy and is able to generalize this ability to produce an appropriate list of features from the co-occurrence vector of a novel concept.     

Word Senses as Clusters of Meaning Modulations: A Computational Model of Polysemy

Most words in natural languages are polysemous; that is, they have related but different meanings in different contexts. This one-to-many mapping of form to meaning presents a challenge to understanding how word meanings are learned, represented, and processed. Previous work has focused on solutions in which multiple static semantic representations are linked to a single word form, which fails to capture important generalizations about how polysemous words are used; in particular, the graded nature of polysemous senses, and the flexibility and regularity of polysemy use. We provide a novel view of how polysemous words are represented and processed, focusing on how meaning is modulated by context. Our theory is implemented within a recurrent neural network that learns distributional information through exposure to a large and representative corpus of English. Clusters of meaning emerge from how the model processes individual word forms. In keeping with distributional theories of semantics, we suggest word meanings are generalized from contexts of different word tokens, with polysemy emerging as multiple clusters of contextually modulated meanings. We validate our results against a human-annotated corpus of polysemy focusing on the gradedness, flexibility, and regularity of polysemous sense individuation, as well as behavioral findings of offline sense relatedness ratings and online sentence processing. The results provide novel insights into how polysemy emerges from contextual processing of word meaning from both a theoretical and computational point of view.     

Modelling the effects of semantic ambiguity in word recognition

Most words in English are ambiguous between different interpretations; words can mean different things in different contexts. We investigate the implications of different types of semantic ambiguity for connectionist models of word recognition. We present a model in which there is competition to activate distributed semantic representations. The model performs well on the task of retrieving the different meanings of ambiguous words, and is able to simulate data reported by Rodd, Gaskell, and Marslen-Wilson [J. Mem. Lang. 46 (2002) 245] on how semantic ambiguity affects lexical decision performance. In particular, the network shows a disadvantage for words with multiple unrelated meanings (e.g., bark) that coexists with a benefit for words with multiple related word senses (e.g., twist). The ambiguity disadvantage arises because of interference between the different meanings, while the sense benefit arises because of differences in the structure of the attractor basins formed during learning. Words with few senses develop deep, narrow attractor basins, while words with many senses develop shallow, broad basins. We conclude that the mental representations of word meanings can be modelled as stable states within a high-dimensional semantic space, and that variations in the meanings of words shape the landscape of this space.     

情绪名词的具体性效应：来自ERP的证据

词汇的具体性和情绪性是影响词汇加工的不同因素, 高具体性和情绪性都能够促进词汇的加工。本研究同时操纵词汇的具体性和情绪性, 使用词汇判定任务和愉悦度判断任务, 探讨了情绪名词的具体性效应及其具体性效应是否受词汇情绪信息的影响。结果发现：情绪名词的具体性效应受内隐或外显情绪条件的影响, 具体的情绪词比抽象的情绪词反应时间更短、正确率更高, 诱发了更大的N400和减小的LPC, 但LPC的具体性效应只表现在内隐情绪任务中。词汇的具体性和情绪性的相互影响发生在内隐情绪任务中的语义加工阶段, 正性、负性的具体词和抽象词的加工在N400成分上差异不显著, 而中性具体词和抽象词在N400成分上差异显著, 说明词汇的情绪信息为抽象词的加工提供了充分的语境, 因此消除了具体词的加工优势。     

Is emotional content obtained from parafoveal words during reading? An eye movement analysis

An eye-movement-contingent display change technique was employed to study whether adult readers extract semantic information from parafoveal words during reading. Three types of parafoveal preview conditions were contrasted: an emotional word, a neutral word, and an identical word condition. To have a maximally effective parafoveal manipulation, high-arousal emotional words (sex- and threat-related and curse words) were used as parafoveal previews. Readers' eye fixation patterns around the target word revealed no evidence for parafoveal semantic processing. Furthermore, the pupil size showed no signs for an emotional response triggered by an emotional word previewed parafoveally. These results are consistent with the view that, as a rule, only the fixated word is processed to a semantic level during reading.     

Primed picture naming within and across languages: An ERP investigation

In two experiments, while event-related potentials (ERPs) were recorded, participants named picture targets that were preceded by masked word primes that corresponded either to the name of the picture target or to an unrelated picture name. Experiment 1 showed significant priming effects in the ERP waveforms, free from articulator artifact, starting as early as 200 msec post target onset. Possible loci of these priming effects were proposed within the framework of generic interactive activation models of word recognition and picture naming. These were grouped into three main components: object-specific structural representations, amodal semantic representations, and word-specific phonological and articulatory representations. Experiment 2 provided an initial test of the possible role of each of these components by comparing within-language repetition priming with priming from translation equivalents in bilingual participants. The early and widespread effects of noncognate translation primes in L1 on picture naming in L2 point to object-specific and amodal semantic representations as the principal loci of priming effects obtained with masked word primes and picture targets.     

Simple Co‐Occurrence Statistics Reproducibly Predict Association Ratings

What determines human ratings of association? We planned this paper as a test for association strength (AS) that is derived from the log likelihood that two words co‐occur significantly more often together in sentences than is expected from their single word frequencies. We also investigated the moderately correlated interactions of word frequency, emotional valence, arousal, and imageability of both words (r's ≤ .3). In three studies, linear mixed effects models revealed that AS and valence reproducibly account for variance in the human ratings. To understand further correlated predictors, we conducted a hierarchical cluster analysis and examined the predictors of four clusters in competitive analyses: Only AS and word2vec skip‐gram cosine distances reproducibly accounted for variance in all three studies. The other predictors of the first cluster (number of common associates, (positive) point‐wise mutual information, and word2vec CBOW cosine) did not reproducibly explain further variance. The same was true for the second cluster (word frequency and arousal); the third cluster (emotional valence and imageability); and the fourth cluster (consisting of joint frequency only). Finally, we discuss emotional valence as an important dimension of semantic space. Our results suggest that a simple definition of syntagmatic word contiguity (AS) and a paradigmatic measure of semantic similarity (skip‐gram cosine) provide the most general performance‐independent explanation of association ratings.     

神经重用假说的汉语情绪词汇加工证据

神经重用假说认为对词的情绪效应可以不经语义形成而先行产生, 这更有助于人类适应环境。为了检验这一假说, 设计脑电实验和行为实验, 记录被试加工汉语厌恶情绪词和中性词的特征。结果发现, 170 ms左右厌恶词与中性词的ERPs发生分离, 且在枕部出现情绪刺激视觉注意关联的EPN; 厌恶词与中性词N400差异波溯源定位于脑岛附近, 其最大激活点时间窗口开始于380 ms。这表明在情绪词汇的语义分析之前就出现了情绪效应, 结果支持了神经重用假说。     

Representing word meaning and order information in a composite holographic lexicon

The authors present a computational model that builds a holographic lexicon representing both word meaning and word order from unsupervised experience with natural language. The model uses simple convolution and superposition mechanisms (cf. B. B. Murdock, 1982) to learn distributed holographic representations for words. The structure of the resulting lexicon can account for empirical data from classic experiments studying semantic typicality, categorization, priming, and semantic constraint in sentence completions. Furthermore, order information can be retrieved from the holographic representations, allowing the model to account for limited word transitions without the need for built-in transition rules. The model demonstrates that a broad range of psychological data can be accounted for directly from the structure of lexical representations learned in this way, without the need for complexity to be built into either the processing mechanisms or the representations. The holographic representations are an appropriate knowledge representation to be used by higher order models of language comprehension, relieving the complexity required at the higher level.     

Linking somatic and symbolic representation in semantic memory: the dynamic multilevel reactivation framework

Biological plausibility is an essential constraint for any viable model of semantic memory. Yet, we have only the most rudimentary understanding of how the human brain conducts abstract symbolic transformations that underlie word and object meaning. Neuroscience has evolved a sophisticated arsenal of techniques for elucidating the architecture of conceptual representation. Nevertheless, theoretical convergence remains elusive. Here we describe several contrastive approaches to the organization of semantic knowledge, and in turn we offer our own perspective on two recurring questions in semantic memory research: (1) to what extent are conceptual representations mediated by sensorimotor knowledge (i.e., to what degree is semantic memory embodied)? (2) How might an embodied semantic system represent abstract concepts such as modularity, symbol, or proposition? To address these questions, we review the merits of sensorimotor (i.e., embodied) and amodal (i.e., disembodied) semantic theories and address the neurobiological constraints underlying each. We conclude that the shortcomings of both perspectives in their extreme forms necessitate a hybrid middle ground. We accordingly propose the Dynamic Multilevel Reactivation Framework—an integrative model predicated upon flexible interplay between sensorimotor and amodal symbolic representations mediated by multiple cortical hubs. We discuss applications of the dynamic multilevel reactivation framework to abstract and concrete concept representation and describe how a multidimensional conceptual topography based on emotion, sensation, and magnitude can successfully frame a semantic space containing meanings for both abstract and concrete words. The consideration of ‘abstract conceptual features’ does not diminish the role of logical and/or executive processing in activating, manipulating and using information stored in conceptual representations. Rather, it proposes that the materials upon which these processes operate necessarily combine pure sensorimotor information and higher-order cognitive dimensions involved in symbolic representation.     

音乐对情感词加工的情感/意义双启动效应

本研究在前人研究的基础上进一步探讨音乐对情感词的启动机制。采用内隐情感启动范式,运用脑电技术考察多种情感类型的音乐片段对情感意义一致或不一致的词加工的启动影响。行为实验结果显示,音乐对情感词产生启动效应。脑电实验的结果发现,音乐对情感词的启动效应表现为早晚两个子成分的N400效应,首先是音乐直接引发的情绪产生第一个启动,表现为早成分的N400;其次是音乐通过联想引发的情感语义而产生第二个启动,表现为晚成分的N400。据此可以认为,音乐对情感词的启动是双启动模式。     

Abstract Conceptual Feature Ratings Predict Gaze Within Written Word Arrays: Evidence From a Visual Wor(l)d Paradigm

The Abstract Conceptual Feature (ACF) framework predicts that word meaning is represented within a high‐dimensional semantic space bounded by weighted contributions of perceptual, affective, and encyclopedic information. The ACF, like latent semantic analysis, is amenable to distance metrics between any two words. We applied predictions of the ACF framework to abstract words using eyetracking via an adaptation of the classical “visual word paradigm” (VWP). Healthy adults (n = 20) selected the lexical item most related to a probe word in a 4‐item written word array comprising the target and three distractors. The relation between the probe and each of the four words was determined using the semantic distance metrics derived from ACF ratings. Eye movement data indicated that the word that was most semantically related to the probe received more and longer fixations relative to distractors. Importantly, in sets where participants did not provide an overt behavioral response, the fixation rates were nonetheless significantly higher for targets than distractors, closely resembling trials where an expected response was given. Furthermore, ACF ratings which are based on individual words predicted eye fixation metrics of probe‐target similarity at least as well as latent semantic analysis ratings which are based on word co‐occurrence. The results provide further validation of Euclidean distance metrics derived from ACF ratings as a measure of one facet of the semantic relatedness of abstract words and suggest that they represent a reasonable approximation of the organization of abstract conceptual space. The data are also compatible with the broad notion that multiple sources of information (not restricted to sensorimotor and emotion information) shape the organization of abstract concepts. While the adapted “VWP” is potentially a more metacognitive task than the classical visual world paradigm, we argue that it offers potential utility for studying abstract word comprehension.     

Hemispheric asymmetries in the split-fovea model of semantic processing

We report a series of neural network models of semantic processing of single English words in the left and the right hemispheres of the brain. We implement the foveal splitting of the visual field and assess the influence of this splitting on a mapping from orthography to semantic representations in single word reading. The models were trained on English four-letter words, presented according to their frequency in all positions encountered during normal reading. The architecture of the model interacted with the training set to produce processing asymmetries comparable to those found in behavioral studies. First, the cueing effects of dominant and subordinate meanings of ambiguous words were different for words presented to the left or to the right of the input layer. Second, priming effects of groups of related words were stronger in the left input than the right input of the model. These effects were caused by coarser-coding in the right half compared with the left half of the model, an emergent effect of the split model interacting with informational asymmetries in the left and right parts of words in the lexicon of English. Some or all of the behavioral data for reading single words in English may have a similar origin.     

Parental Separation and Children's Behavioral/Emotional Problems: The Impact of Parental Representations and Family Conflict

In this longitudinal study, we examine whether the effect of parental separation on kindergarten children's behavioral/emotional problems varies according to the level of family conflict, and children's parental representations. One hundred and eighty seven children were assessed at ages 5 and 6. Family conflict was assessed using parents' ratings. Children's parental representations were assessed using a story-stem task. A multiinformant approach (parent, teacher, child) was employed to assess children's behavioral/emotional problems. Bivariate results showed that separation, family conflict, and negative parental representations were associated with children's behavioral/emotional problems. However, in multivariate analyses, when controlling for gender and symptoms at age 5, we found that children of separated parents who showed negative parental representations had a significantly greater increase in conduct problems between 5 and 6 than all other children. In terms of emotional symptoms and hyperactivity, symptoms at 5 and (for hyperactivity only) gender were the only predictors for symptoms 1 year later. Our results suggest that kindergarten children's representations of parent-child relationships moderate the impact of parental separation on the development of conduct problems, and underline play and narration as a possible route to access the thoughts and feelings of young children faced with parental separation.     

A synchronization account of false recognition

We describe a computational model to explain a variety of results in both standard and false recognition. A key attribute of the model is that it uses plausible semantic representations for words, built through exposure to a linguistic corpus. A study list is encoded in the model as a gist trace, similar to the proposal of fuzzy trace theory (Brainerd & Reyna, 2002), but based on realistically structured semantic representations of the component words. The model uses a decision process based on the principles of neural synchronization and information accumulation. The decision process operates by synchronizing a probe with the gist trace of a study context, allowing information to be accumulated about whether the word did or did not occur on the study list, and the efficiency of synchronization determines recognition. We demonstrate that the model is capable of accounting for standard recognition results that are challenging for classic global memory models, and can also explain a wide variety of false recognition effects and make item-specific predictions for critical lures. The model demonstrates that both standard and false recognition results may be explained within a single formal framework by integrating realistic representation assumptions with a simple processing mechanism.     

Talking About Past Emotions: Conversations Between Peruvian Mothers and Their Preschool Children

In this study, we examined gender and age variations in the use of emotion words during mother–child conversations about experiences. Thirty-two middle-class, Spanish-speaking, Peruvian mothers and their 3- and 5-year-old children participated in this study. Conversations were audiotaped, transcribed, and coded for number, types of emotions discussed, and conversational contexts in which the emotion word was embedded. Age differences were found only in children's use of emotion words. Gender differences were found only in mothers' use of emotion words. Findings are discussed in relation to socialization practices of emotional expression and gender in middle-class Peruvian society. Our results support and provide a developmental perspective on findings obtained in studies of adult emotional expression across various cultures.     

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.     

Behavioral and neural analyses of connotative meaning: Word classes and rating scales

Behavioral (semantic differential) and neural (Evoked Potentials, EPs) responses were related to connotative meaning. The approach was based on Osgood's semantic analyses and dimensions of Evaluation (E), Potency (P), and Activity (A). The experimental variables were (1) the semantic class of the stimulus word (E+, E-, P+, P-, A+, A-) and (2) the dimension of the semantic scale (E, P, A) which the subject used to rate the stimulus words. These variables were experimentally combined such that on each trial the subject used a designated semantic scale to judge a specified stimulus word while brain activity was recorded. Using multivariate analyses, the effects on the EPs of stimulus word class, scale dimension, and their interaction were analyzed. The EP effects of stimulus word class were similar whether the subjects were merely saying the words or rating the words on a variety of semantic scales. Different EPs were found for six word classes, three semantic scale dimensions, and the 18 groups formed by their combination. The success rates in EP identification of (1) word class and (2) scale dimension did not depend on whether these two kinds of semantic variables involved the same or different semantic dimensions. The two kinds of semantic effects in EPs were largely independent. The behavioral data supported Osgood's results and showed that our subjects were appropriately processing the semantic information. The common analyses of data from all subjects suggest the universality of the connotative EP effects across individuals. This parallels, at the neural level, the universality of the connotative dimensions found at the behavioral level by semantic differential ratings. The EP effects imply that the neural representation of meaning is similar in different individuals.