Canada

Dementia is a common and debilitating condition that typically gives rise to increasing language impairment. There is a need to understand the nature of this impairment further so that therapies may be developed, particularly in the case of bilinguals. This paper extends BiLex, an existing computational model of bilingual lexical access, to simulate language decline in dementia. Six lesion types are evaluated for their ability to reproduce the pattern of decline in the semantic variant primary progressive aphasia (svPPA) subtype of dementia. Semantic memory lesions reproduce this pattern of decline best in monolinguals, and further suggest patterns that are likely to be found in longitudinal data from bilingual dementia patients in the future.

CogSci 2022

Constructing Individualized Computational Models for Dementia Patients

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

poster

Join us for CogSci2022! Our 44th Annual Conference will be centered around the theme of diversity, with an accompanying range of keynotes and invited speakers who themselves highlight the diversity of our field. Our three symposia are Cognition across Diverse Populations, Cognition within Diverse Societies, and Global Challenges for Cognitive Science. The conference is being offered in a fully hybrid environment for onsite and online attendees.
 
 [**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/07/Cogsci-2022-PAAG-Detailed-July-12final-1.pdf)
 [**The Detailed Program can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/08/DeArmond-CogSci2022-program-v113-final.pdf)

This conference requires a registration to attend. Please use the link below to secure your ticket. The website will be available to all registered attendees on July 25th. 

[Register to Attend](https://cognitivesciencesociety.org/registration/)

Registration Is Required

The 44th Annual Meeting of the Cognitive Science Society highlighted research on the theme Cognitive Diversity in addition to the full breadth and diversity of research topics offered by the society’s membership. 





The effect of prioritization on information in working memory
has primarily been examined in tasks containing a single type
of stimulus and with one item that is prioritized. However,
many theories of working memory posit different types of
components for the maintenance of verbal or visuospatial
information. This study examined differences between
prioritized and nonprioritized items as well as word and image
stimuli. Participants completed an association learning task in
which working memory demands were varied along with the
number of items to be prioritized. Following a short delay,
retention was tested. Prioritization effects were identified
during both the learning and testing phases of the experiment,
and the impact of prioritization was moderated by working
memory demands of the task. Significant differences in
accuracy between word and image stimuli were only observed
in the testing phase, with accuracy for verbal information being
worse. While prioritization improved accuracy and response
times during learning, it led to decreases in the testing phase.

Examining Prioritization in Working Memory for Verbal and Visual Stimuli

Visual working memory and visual mental imagery both involve the use of internal visual representations, and they likely have overlapping neural substrates. However, research on people with “aphantasia,” or a lack of visual imagery, has not found any evidence that aphantasics are impaired on visual working memory tasks, possibly because they can use non-visual strategies. We designed a task intended to prevent compensatory strategies, and also to explore what happens when aphantasics are required to shift the focus of attention between items in working memory. We found that aphantasics were not significantly different from controls, either when maintaining or shifting the focus of attention. Explanations include non-visual memory strategies, but also the possibility that aphantasics can store information in visual working memory without conscious awareness. Future research should combine behavioral methods with neuroimaging to investigate how aphantasics encode working memory representations.


Memory without Imagery: No Evidence of Visual Working Memory Impairment in People with Aphantasia

A critical part of language comprehension is inferring omitted but plausible information from linguistic descriptions of events. For instance, the verb phrase ‘preserve vegetable’ implies the instrument vinegar whereas ‘preserve fruit’ implies dehydrator. We studied the ability of distributional semantic models to perform this kind of semantic inference after being trained on an artificial corpus with strictly controlled constraints on which verb phrases occur with which instruments. Importantly, the ability to infer omitted but plausible instruments in our task requires compositional generalization. We found that contemporary neural network models fall short generalizing learned selectional constraints, and that a graph-based distributional semantic model trained on constituency-parsed data and equipped with a spreading-activation procedure for calculating semantic relatedness, achieves perfect performance. Our findings shed light on the mechanisms that give rise to compositional generalization, and using graphs to model semantic memory.

Compositional Generalization in a Graph-based Model of Distributional Semantics

Learning to identify a person’s voice is a key component of speech perception. In this study, we use a categorization framework to provide insights about the mechanisms supporting talker identification. Native Mandarin Chinese listeners learned to categorize sentences in three tasks with different language contexts – native Mandarin talkers speaking Mandarin, native English talkers speaking English, and native Mandarin talkers speaking English. We compared learning when listeners received fully informative or minimal feedback. Using decision bound models, we examined the strategies participants used in each of the three tasks. Regardless of language context, full feedback was initially better for learning than minimal feedback but was no different after the second block. Across tasks, participants often used strategies based on mean fundamental frequency to separate the talkers. These results demonstrate that talker identification is a categorization problem, which enables leveraging existing category learning frameworks to understand the mechanisms of this important ability.

Talker identification as a categorization problem

Categorical perception (CP) facilitates the discrimination of stimuli belonging to different categories relative to those from the same category. Effects of CP on the discrimination of color and shape have been attributed to the top-down modulation of visual perception by the left-lateralized language processes. We used a divided visual field (DVF) search paradigm to investigate the prospective effects of CP for face identity and gender processing. Consistent with visual processing of face identity in the right hemisphere, we found CP facilitated perception only in the left visual field (LVF). In contrast, and consistent with language-induced CP, we observed a between-category advantage for processing face gender only in the right visual field (RVF). Taken together, our results suggest that language-induced CP plays a role in the category-based visual processing of faces by the left hemisphere, but face familiarity processing might be dependent on different, identity-specialized networks in the right hemisphere.

Language-induced categorical perception of faces?

The presence of semantic information in multivariate patterns of neural activity has been explored as a method of measuring knowledge and learning. Using fMRI, we investigated whether novice learners of American Sign Language (ASL) showed overlapping representations of semantic categories for words presented in a well-known (English) or newly learned (ASL) language. We find evidence of neural patterns that were partially shared between sign and speech in novice participants. This result provides evidence for the influence of even brief learning on neural representations in cross-modality language processing.

Overlapping semantic representations of sign and speech in novice sign language learners

This study focuses on naming and rating ambiguously categorized objects to more accurately represent the real-life naming scenarios that language-users encounter. We examined 407 images of common object categories from a previous study collected in 2013 with monolingual undergraduate students in the United States and China. Of those images, we selected 150 that both (1) balanced the mean and range of naming variables between languages and (2) had low between-language correlations on the same variables. We elicited object names and typicality ratings for these 150 images from a more age-, education-, and geographically-diverse sample of English monolinguals across the US. Naming measures were significantly correlated between current and previous English samples, suggesting high consistency of these object categories within language, especially typicality ratings when conditioned on the object name typicality is not a strictly conceptual measure. Cross-language correlations were low, illustrating the unique categorization patterns between languages.

Lexical Categorization Variables for English and Mandarin Quantify Ambiguity in Object Naming

Memory is limited in capacity, which means that we must choose what information to prioritize for storage. Part of knowing what to prioritize is predicting future needs. For example, if you view a 3D object, later on you may wish to recall exactly how it was oriented. Alternatively, you might need to remember its shape, independent of viewpoint. Given this kind of uncertainty, a good strategy would be to store multiple kinds of information about the objects we observe, and then decode in a task-dependent manner. We tested whether people apply these strategies in the specific domain of short-term memory for novel faces. To test whether people store various kinds of information about a face, and then decode in a task-dependent manner, we modeled their responses in a memory task using features (extracted from deep neural networks) that varied in how much 3D information they carried. We found strong evidence for a mixed-storage strategy, which did not vary in response to task demands. Our results suggest that in order to fully understand resource allocation and retrieval strategies in human memory, it may be critical to consider not just the distribution over tasks in people's natural environments, but also task uncertainty at the time of encoding.

Coding Strategies in Memory for 3D Objects: The Influence of Task Uncertainty

Recent work in judgment and decision making has focused on which actions people consider when solving open-ended problems and found that the actions that come to mind tend to have particular features, such as having a high historical value. Here, we pursue the idea that the process of generating actions for decision-making tasks may actually reflect more general mechanisms for generating kinds of things. We provide evidence that what comes to mind may simply be a reflection of participants sampling from the most relevant part of the representational space they use to encode the type of thing they are generating. In this paper, we (1) introduce an approach for empirically describing a category in terms of the features that people use to represent category members, and for locating category members within that feature space, (2) show that certain locations in a category's feature space predict an item's likelihood of coming to mind, (3) introduce an approach for understanding the relevance of various features to people's representations of category members, and (4) show that features which are most involved in people's representations of category members are also predictors of what comes to mind within a category. We close by proposing that features that are most relevant to our representations of category members and predict coming to mind are those for which it has been historically useful to have information about during past experiences with the category in question.

What comes to mind? Samples from relevance-based feature spaces

Testing the testing effect with featural and relational categories
Theoreticians have recently proposed that the benefits of testing depend on the extent to which the to-be-learned concepts are interconnected, with testing effects emerging from low element interconnectivity and becoming weaker with greater element interconnectivity. We tested this idea by using a classification task and manipulating the amount of element interconnectivity in the categories that subjects learned. Some subjects learned featural categories, which lack element interconnectivity, whereas others learned relational categories, which are defined by how its elements are interconnected. This factor was crossed with type of training, wherein some subjects learned through classification (i.e., testing) and others through observation (i.e., studying). For featural categories, it was predicted that classification training would produce better learning than observational training, whereas this benefit should become weaker or disappear for relational categories. However, preliminary results show that both types of training lead to similar learning for both category types.

Premium content

Downloads

Next from CogSci 2022

Examining Prioritization in Working Memory for Verbal and Visual Stimuli

Similar lecture

Adversarial Text Generation using Large Language Models for Dementia Detection