Models Of Semantic Memory Research Articles

Passenger Behavior Prediction With Semantic and Multi-Pattern LSTM Model

Understanding passenger behaviors is of great importance in intelligent transportation and infrastructure planning. However, the passenger trajectories are actually complex temporal data, which consist of rich spatial and temporal information. What’s more, the observed passenger trajectories may be a mixture of different types of passengers with various travelling purposes. These difficulties make the prediction of passenger behaviors a challenging work. To address these problems, this paper improves the existing passenger behavior prediction methods from the following two aspects: 1) Encoding the travelling sequence with personalized semantic sensing, and 2) constructing multi-pattern prediction models to capture multiple travelling purposes and dynamics. Along this line, this paper provides a novel passenger behavior prediction model, namely, Semantic and multi-Pattern Long Short-Term Memory (SP-LSTM) model. Particularly, 1) a translation unit is designed, which is able to encode an observed travelling sequence into a structured sequence with consideration of individual travelling semantics; 2) a multi-pattern learning schematic is proposed, which first identifies the travelling patterns of passengers and then handles different patterns with different learning modules; 3) a unified learning framework is provided to integrate the semantic sensing module and multi-pattern learning module together, and present the final prediction results. To evaluate the proposed method, this paper conducts experiments on real-world passenger travelling data. Results demonstrate the superiority of SP-LSTM over both classical and the state-of-the-art methods.

Distinctive semantic features in the healthy adult brain.

The role of semantic features, which are distinctive (e.g., a zebra's stripes) or shared (e.g. has four legs) for accessing a concept, has been studied in detail in early neurodegenerative disease such as semantic dementia (SD). However, potential neural underpinnings of such processing have not been studied in healthy adults. The current study examines neural activation patterns using fMRI while participants completed a feature verification task, in which they identified shared or distinctive semantic features for a set of natural kinds and man-made artifacts. The results showed that the anterior temporal lobe bilaterally is an important area for processing distinctive features, and that this effect is stronger within natural kinds than man-made artifacts. These findings provide converging evidence from healthy adults that is consistent with SD research, and support a model of semantic memory in which patterns of specificity of semantic information can partially explain differences in neural activation between categories.

Association and response accuracy in the wild.

We studied contestant accuracy and error in a popular television quiz show, "Jeopardy!" Using vector-based knowledge representations obtained from distributional models of semantic memory, we computed the strength of association between clues and responses in over 5,000 televised games. Such representations have been shown to play a key role in memory and judgment, and consistent with this work, we find that contestants are more likely to provide correct responses when these responses are strongly associated with their clues, and more likely to provide incorrect responses when correct responses are weakly or negatively associated with their clues. This effect is stronger for easier questions with low monetary values, and for questions in which contestants compete to respond quickly. Our results show how distributional models of semantic memory can be used to predict human behavior in naturalistic high-level judgment tasks with skilled participants and significant monetary and social incentives.

A feature-based neurocomputational model of semantic memory.

According with a featural organization of semantic memory, this work is aimed at investigating, through an attractor network, the role of different kinds of features in the representation of concepts, both in normal and neurodegenerative conditions. We implemented new synaptic learning rules in order to take into account the role of partially shared features and of distinctive features with different saliency. The model includes semantic and lexical layers, coding, respectively for object features and word-forms. Connections among nodes are strongly asymmetrical. To account for the feature saliency, asymmetrical synapses were created using Hebbian rules of potentiation and depotentiation, setting different pre-synaptic and post-synaptic thresholds. A variable post-synaptic threshold, which automatically changed to reflect the feature frequency in different concepts (i.e., how many concepts share a feature), was used to account for partially shared features. The trained network solved naming tasks and word recognition tasks very well, exploiting the different role of salient versus marginal features in concept identification. In the case of damage, superordinate concepts were preserved better than the subordinate ones. Interestingly, the degradation of salient features, but not of marginal ones, prevented object identification. The model suggests that Hebbian rules, with adjustable post-synaptic thresholds, can provide a reliable semantic representation of objects exploiting the statistics of input features.

Trait Associations for Hillary Clinton and Donald Trump in News Media

We study media representations of Hillary Clinton and Donald Trump in the 2016 U.S. presidential election. In particular, we train models of semantic memory on a large number of news media outlets that published online articles during the course of the election. Based on the structure of word co-occurrence in these media outlets, our models learn semantic representations for the two presidential candidates as well as for widely studied personality traits. We find that models trained on media outlets most read by Clinton voters and media outlets most read by Trump voters differ in the strength of association between the two candidates’ names and trait words pertaining to morality. We observe some differences for trait words pertaining to warmth, but none for trait words pertaining to competence.

Associative Agreement as a Predictor of Naming Ability in Alzheimer's Disease: A Case for the Semantic Nature of Associative Links.

We aimed to address the long-standing issue of the nature of the relationships that link a cue word to words associated with it. In keeping with a recently proposed neuropsychological model of semantic memory (Zannino et al., 2015), we provide support for the hypothesis that associative links are semantic in nature and not lexical. In support of this hypothesis, we demonstrate a relationship in healthy subjects between the probability of producing word X in response to cue word Y in a free association task and the probability of using word X to describe the meaning of word Y. Furthermore, we provide evidence that associative measures are altered in people suffering from Alzheimer's disease (AD) and predict their level of performance in a picture-naming task. We provide a parsimonious account of the experimental data gathered form these different sources of evidence according to the hypothesis that the links between a cue word and its associates can be viewed as binding a concept (the cue) to pieces of information regarding its meaning (the associates).

Towards a unified model of semantic memory: validation and theoretical implications of the conceptual feature rating space

ABSTRACTThe representation of abstract and concrete concepts has long been a debated question in the literature. In previous work, we created a semantic space which was able to model abstract and concrete concepts in a single multidimensional space. In this study, we tested the validity of our semantic space by comparing it with Latent Semantic Analysis in a word relatedness task during eyetracking. We measured reaction time, gaze measures, and pupil dilation. The results of our study revealed that the newly created space was valid and not just a product of our data. Our ability to simulate a high dimensional semantic space represents a step forward in our understanding of the representation of concepts.

Integration of Semantic and Episodic Memories.

This paper describes the integration of semantic and episodic memory (EM) models and the benefits of such integration. Semantic memory (SM) is used as a foundation of knowledge and concept learning, and is needed for the operation of any cognitive system. EM retains personal experiences stored based on their significance-it is supported by the SM, and in return, it supports SM operations. Integrated declarative memories are critical for cognitive system development, yet very little research has been done to develop their computational models. We considered structural self-organization of both semantic and episodic memories with a symbolic representation of input events. Sequences of events are stored in EM and are used to build associations in SM. We demonstrated that integration of semantic and episodic memories improves the native operation of both types of memories. Experimental results are presented to illustrate how the two memories complement each other by improving recognition, prediction, and context-based generalization of individual memories.

The semantic representation of prejudice and stereotypes

We use a theory of semantic representation to study prejudice and stereotyping. Particularly, we consider large datasets of newspaper articles published in the United States, and apply latent semantic analysis (LSA), a prominent model of human semantic memory, to these datasets to learn representations for common male and female, White, African American, and Latino names. LSA performs a singular value decomposition on word distribution statistics in order to recover word vector representations, and we find that our recovered representations display the types of biases observed in human participants using tasks such as the implicit association test. Importantly, these biases are strongest for vector representations with moderate dimensionality, and weaken or disappear for representations with very high or very low dimensionality. Moderate dimensional LSA models are also the best at learning race, ethnicity, and gender-based categories, suggesting that social category knowledge, acquired through dimensionality reduction on word distribution statistics, can facilitate prejudiced and stereotyped associations.

Application of Connectionist Semantic Memory Model in Building QandA Bank

Application of Connectionist Semantic Memory Model in Building QandA Bank

The relational luring effect: Retrieval of relational information during associative recognition.

Here we argue that semantic relations (e.g., works in: nurse-hospital) have abstract independent representations in long-term memory (LTM) and that the same representation is accessed by all exemplars of a specific relation. We present evidence from 2 associative recognition experiments that uncovered a novel relational luring effect (RLE) in recognition memory. Participants studied word pairs, and then discriminated between intact (old) pairs and recombined lures. In the first experiment participants responded more slowly to lures that were relationally similar (table-cloth) to studied pairs (floor-carpet), in contrast to relationally dissimilar lures (pipe-water). Experiment 2 extended the RLE by showing a continuous effect of relational lure strength on recognition times (RTs), false alarms, and hits. It used a continuous pair recognition task, where each recombined lure or target could be preceded by 0, 1, 2, 3 or 4 different exemplars of the same relation. RTs and false alarms increased linearly with the number of different previously seen relationally similar pairs. Moreover, more typical exemplars of a given relation lead to a stronger RLE. Finally, hits for intact pairs also rose with the number of previously studied different relational instances. These results suggest that semantic relations exist as independent representations in LTM and that during associative recognition these representations can be a spurious source of familiarity. We discuss the implications of the RLE for current models of semantic and episodic memory, unitization in associative recognition, analogical reasoning and retrieval, as well as constructive memory research. (PsycINFO Database Record

Lexical-semantic deficits in processing food and non-food items

The study of category specific deficits in brain-damaged patients has been instrumental in explaining how knowledge about different types of objects is organized in the brain. Much of this research focused on testing putative semantic sensory/functional subsystems that could explain the observed dissociations in performance between living things (e.g., animals and fruits/vegetables) and non-living things (e.g., tools). As neuropsychological patterns that did not fit the original living/non-living distinction were observed, an alternative organization of semantic memory in domains constrained by evolutionary pressure was hypothesized. However, the category of food, that contains both living-natural items, such as an apple, and nonliving-manufactured items as in the case of a hamburger, has never been systematically investigated. As such, food category could turn out to be very useful to test whether the brain organizes the knowledge about food in sensory/functional subsystems, in a specific domain, or whether both approaches might need to be integrated. In the present study we tested the ability of patients with Alzheimer dementia (AD) and with Primary Progressive Aphasias (PPA) as well as healthy controls to perform a confrontation naming task, a categorization task, and a comprehension of edible (natural and manufactured food) and non edible items (tools and non-edible natural things) task (Tasks 1–3). The same photographs of natural and manufactured food were presented together with a description of food’s sensory or functional property that could be either congruent or incongruent with that particular food (Task 4). Patients were overall less accurate than healthy individuals, and PPA patients were generally more impaired than AD patients, especially on the naming task. Food tended to be processed better than non-food in two out of three tasks (categorization and comprehension tasks). Patient groups showed no difference in naming food and non-food items, while controls were more accurate with non-food than food (controlling for the linguistic variables and calorie content). AD patients named manufactured food more accurately than natural food (with PPA and controls showing no difference). Recognition of food and, to some extent, of manufactured food seems to be more resilient to brain damage, possibly by virtue of its survival relevance. Furthermore, on Task 4 patients showed an advantage for the sensory-natural pairs over sensory-manufactured combination. Overall, findings do not fit an existing model of semantic memory and suggest that properties intrinsic to the food items (such as the level of transformation and the calorie content) or even to the participants like the Body Mass Index (as shown in another study reviewed here) should be considered.

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 predicatedupon 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.

Does the sound of a barking dog activate its corresponding visual form? An fMRI investigation of modality-specific semantic access

Much remains to be learned about the neural architecture underlying word meaning. Fully distributed models of semantic memory predict that the sound of a barking dog will conjointly engage a network of distributed sensorimotor spokes. An alternative framework holds that modality-specific features additionally converge within transmodal hubs. Participants underwent functional MRI while covertly naming familiar objects versus newly learned novel objects from only one of their constituent semantic features (visual form, characteristic sound, or point-light motion representation). Relative to the novel object baseline, familiar concepts elicited greater activation within association regions specific to the presentation modality. Furthermore, visual form elicited activation within high-level auditory association cortex. Conversely, environmental sounds elicited activation in regions proximal to visual association cortex. Both conditions commonly engaged a putative hub region within lateral anterior temporal cortex. These results support hybrid semantic models in which local hubs and distributed spokes are dually engaged in service of semantic memory.

Lexical processing and organization in bilingual first language acquisition: Guiding future research.

A rich body of work in adult bilinguals documents an interconnected lexical network across languages, such that early word retrieval is language independent. This literature has yielded a number of influential models of bilingual semantic memory. However, extant models provide limited predictions about the emergence of lexical organization in bilingual first language acquisition (BFLA). Empirical evidence from monolingual infants suggests that lexical networks emerge early in development as children integrate phonological and semantic information. These findings tell us little about the interaction between 2 languages in early bilingual memory. To date, an understanding of when and how languages interact in early bilingual development is lacking. In this literature review, we present research documenting lexical-semantic development across monolingual and bilingual infants. This is followed by a discussion of current models of bilingual language representation and organization and their ability to account for the available empirical evidence. Together, these theoretical and empirical accounts inform and highlight unexplored areas of research and guide future work on early bilingual memory. (PsycINFO Database Record

Graph-Theoretic Properties of Networks Based on Word Association Norms: Implications for Models of Lexical Semantic Memory.

We compared the ability of three different contextual models of lexical semantic memory (BEAGLE, Latent Semantic Analysis, and the Topic model) and of a simple associative model (POC) to predict the properties of semantic networks derived from word association norms. None of the semantic models were able to accurately predict all of the network properties. All three contextual models over-predicted clustering in the norms, whereas the associative model under-predicted clustering. Only a hybrid model that assumed that some of the responses were based on a contextual model and others on an associative network (POC) successfully predicted all of the network properties and predicted a word's top five associates as well as or better than the better of the two constituent models. The results suggest that participants switch between a contextual representation and an associative network when generating free associations. We discuss the role that each of these representations may play in lexical semantic memory. Concordant with recent multicomponent theories of semantic memory, the associative network may encode coordinate relations between concepts (e.g., the relation between pea and bean, or between sparrow and robin), and contextual representations may be used to process information about more abstract concepts.

Generating structure from experience: A retrieval-based model of language processing.

Standard theories of language generally assume that some abstraction of linguistic input is necessary to create higher level representations of linguistic structures (e.g., a grammar). However, the importance of individual experiences with language has recently been emphasized by both usage-based theories (Tomasello, 2003) and grounded and situated theories (e.g., Zwaan & Madden, 2005). Following the usage-based approach, we present a formal exemplar model that stores instances of sentences across a natural language corpus, applying recent advances from models of semantic memory. In this model, an exemplar memory is used to generate expectations about the future structure of sentences, using a mechanism for prediction in language processing (Altmann & Mirković, 2009). The model successfully captures a broad range of behavioral effects-reduced relative clause processing (Reali & Christiansen, 2007), the role of contextual constraint (Rayner & Well, 1996), and event knowledge activation (Ferretti, Kutas, & McRae, 2007), among others. We further demonstrate how perceptual knowledge could be integrated into this exemplar-based framework, with the goal of grounding language processing in perception. Finally, we illustrate how an exemplar memory system could have been used in the cultural evolution of language. The model provides evidence that an impressive amount of language processing may be bottom-up in nature, built on the storage and retrieval of individual linguistic experiences.

Hemispheric lateralization of semantic feature distinctiveness

Recent models of semantic memory propose that the semantic representation of concepts is based, in part, on a network of features. In this view, a feature that is distinctive for an object (a zebra has stripes) is processed differently from a feature that is shared across many objects (a zebra has four legs). The goal of this paper is to determine whether there are hemispheric differences in such processing. In a feature verification task, participants responded ‘yes’ or ‘no’ following concepts which were presented to a single visual field (left or right) paired with a shared or distinctive feature. Both hemispheres showed faster reaction times to shared features than to distinctive features, although right hemisphere responses were significantly slower overall and particularly in the processing of distinctive features. These findings support models of semantic processing in which the dominant left hemisphere more efficiently performs highly discriminating ‘fine’ encoding, in contrast to the right hemisphere which performs less discriminating ‘coarse’ encoding.

Preservation of lexical-semantic knowledge of adjectives in the semantic variant of primary progressive aphasia: Implications for theoretical models of semantic memory

Abstract Adjectives are lexical words used to modify nouns and refer to physical and concrete properties as well as abstract properties denoted by nouns. Although adjectives directly refer to semantic attributes, no study has explicitly examined the comprehension of adjectives in the semantic variant of primary progressive aphasia (svPPA). In this study, we explored the comprehension of adjectival meaning in four participants with svPPA using two experimental designs assessing the comprehension of adjectives per se and the comprehension of adjectives when associated with concrete nouns. We found evidence that participants: a) were unimpaired on tasks probing knowledge about adjectives referring to dimension and physical properties, to human propensity and value, and to color, but b) showed substantial difficulty when asked to associate the same adjectives with object nouns. These results raise important questions about the lexical-semantic representation of adjectives and, more generally, about the organization of words in the lexical-semantic system. The observed difficulties could be due to the impairment in binding different modality-specific properties and combining verbal conceptual knowledge, after anterior temporal lobe atrophy.

Neurocomputational approaches to modelling multisensory integration in the brain: A review

The Brain’s ability to integrate information from different modalities (multisensory integration) is fundamental for accurate sensory experience and efficient interaction with the environment: it enhances detection of external stimuli, disambiguates conflict situations, speeds up responsiveness, facilitates processes of memory retrieval and object recognition. Multisensory integration operates at several brain levels: in subcortical structures (especially the Superior Colliculus), in higher-level associative cortices (e.g., posterior parietal regions), and even in early cortical areas (such as primary cortices) traditionally considered to be purely unisensory. Because of complex non-linear mechanisms of brain integrative phenomena, a key tool for their understanding is represented by neurocomputational models. This review examines different modelling principles and architectures, distinguishing the models on the basis of their aims: (i) Bayesian models based on probabilities and realizing optimal estimator of external cues; (ii) biologically inspired models of multisensory integration in the Superior Colliculus and in the Cortex, both at level of single neuron and network of neurons, with emphasis on physiological mechanisms and architectural schemes; among the latter, some models exhibit synaptic plasticity and reproduce development of integrative capabilities via Hebbian-learning rules or self-organizing maps; (iii) models of semantic memory that implement object meaning as a fusion between sensory–motor features (embodied cognition). This overview paves the way to future challenges, such as reconciling neurophysiological and Bayesian models into a unifying theory, and stimulates upcoming research in both theoretical and applicative domains.