Canada

Geometric analogy problems remain an intriguing part of intelligence scales, which is closely correlated to many cognitive studies, such as perception, conception, memory, abstract and inductive reasoning. The problems not only target the most fundamental element --- analogy-making --- in human cognition, but also require integration of multiple components and stages: looking at the test booklet, thinking for a minute or two, and deciding the answer. Great efforts and achievements have been made to explain different individual aspects of this process. In this paper, we take a more holistic approach from the perspective of problem-solving, by modeling the entire process, from the moment the visual stimuli are received to the moment an answer is decided. Therefore, we explore how the final solution can be built upon visual inputs and necessary components that lie between the perceptual input and conceptual output. Particularly, we designed a novel similarity metric and a correspondence-finding method based on mapping and optimization. With these two basic blocks, we implemented a computational model, and report our initial results on a classical problem set.

CogSci 2022

An End-to-End Imagery-Based Modeling of Solving Geometric Analogy Problems

## Welcome to EMNLP 2022!
I am delighted to welcome you to EMNLP 2022! I believe this conference has been complicated beyond any precedent. Over the past year, it’s been thrilling to see the organization team approach each new puzzle with creativity and enthusiasm. We hope that those participating in Abu Dhabi as well as those joining remotely will leave the conference feeling newly inspired by the program and newly connected to our ever-growing community. Following EMNLP 2021 and major NLP conferences since, EMNLP 2022 is “hybrid,” serving both virtual and in-person participants.

Our key innovations for EMNLP 2022 include:

* EMNLP 2022 is “hybrid” in a second sense, as well: we allowed both direct and rolling review paper submissions, building on the pilot experiment of EMNLP 2021, which considered a small number of ARR submissions. 
* Familiar from NAACL but new to EMNLP, we’ve added an industry track.
* During the conference, “portals” will link virtual poster sessions to in-person conference participants during poster sessions each day.
* The first ACL-family conference in the United Arab Emirates.

 *Message from Noah A. Smith, University of Washington and Allen Institute for AI, Seattle, Washington, USA* 
***EMNLP 2022 General Chair***
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/9eec7d4a287ee18c278b08229290aa83.png)](https://drive.google.com/file/d/1OlPv6QBeo62VVTughj2jkiLeyHd1WnUt/view)
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/a3db7a768409f05192210d98601edb25.png)](https://emnlp2022.rocket.chat/)

To access this site you need to register. Please register [here](https://2022.emnlp.org/registration/).

Register here

EMNLP 2022

Welcome!
EMNLP 2022 will take place in Abu Dhabi from December 7th to December 11th, 2022. And it will be held in hybrid mode, both online and offline.

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. 





In STEM instruction, the effectiveness of teaching by analogy is often limited by students’ focus on superficial features of the source and target exemplars; the strategy of progressive alignment (moving from perceptually similar to different targets) has been suggested to address this issue (Gentner & Hoyos, 2017). In contrast, computational models suggest maximizing surface feature variation to improve the relation learning, and human behavioral studies find a relational bias in learning (Austerweil & Ehrens, 2018). Here, participants were explicitly instructed to match stimuli based on relations while perceptual similarity of stimuli varied parametrically. We found that lower perceptual similarity reduced accurate relational matching (F = 9.53, p<.001), and observed a similar trend for reaction times. This finding demonstrates that perceptual similarity may interfere with relational judgements, but also hints at why progressive alignment may be effective. Implications for instructional sequence design will be discussed.

Perceptual Similarity Affects Relational Judgements

Understanding abstract relations, and reasoning about various instantiations of the same relation, is an important marker in human cognition. Here we focus on development of understanding for the concept of antonymy. We examined whether four- and five-year-olds (N= 67) are able to complete an analogy task involving antonyms, whether language cues facilitate children’s ability to reason about the antonym relation, and how their performance compares with that of two vector-based computational models. We found that explicit relation labels in the form of a relation phrase (“opposites”) improved performance on the task for five-year-olds but not four-year-olds. Five-year-old (but not four-year-old) children were more accurate for adjective and verb antonyms than for noun antonyms. Two computational models showed substantial variability in performance across different lexical classes, and in some cases fell short of children’s accuracy levels. These results suggest that young children acquire a solid understanding of the abstract relation of opposites, and can generalize it to various instantiations across different lexical classes. These developmental results challenge relation models based on vector semantics, and highlight the importance of examining performance across different parts of speech.

Children’s Acquisition of the Concept of Antonym Across Different Lexical Classes

Analogy plays an important role in cognitive development, but children often need cognitive supports to draw correct ones. Here, we examined the role of conceptual knowledge in proportional analogies, which are often depicted as a simple exercise in pattern completion. In Study 1, adults and children (N = 321) completed 4-term analogy tasks featuring letters, lines, integers, or fractions. Performance was lowest for fractions, and strongly impacted by educational background. In Study 2, we conducted an educational intervention focusing on either conceptual knowledge, procedural knowledge, or both for 3rd-to-5th graders (N = 343) using a pretest-training-posttest design. Children with poor pretest magnitude knowledge were more likely to fail analogical reasoning, and training on conceptual knowledge that fractions denote magnitudes improved children’s analogies. Together, these studies indicate that knowledge of fractional magnitudes is important to proportional analogy.

Conceptual Prerequisites for Proportional Analogy

One of the advantages of using metaphorical expressions over literal ones might be that speakers can convey not only the intended property, but also its degree. 
For example, when hearing “John is a shark”, the listener might infer that the speaker aims to communicate that John is as mean as a typical shark. 
We present experimental findings supporting this hypothesis, along with a novel metaphor interpretation model, which is implemented within the Rational Speech Act framework.
We compare our model's predictions to those of an existing RSA model of metaphor understanding, within which the listener infers just the presence or absence of a feature as opposed to its degree, and find that our model produces a significantly better fit.

Pragmatics of Metaphor Revisited: Modeling the Role of Degree and Salience in Metaphor Understanding

Structural priming is a heavily studied and multi-faceted phenomenon. Essentially, when other factors are equal, syntax shows a tendency to repeat across utterances with potential gains in fluency (Pickering & Ferreira, 2008). Lexical repetition, particularly repetition of the verb between utterances enhances structural priming significantly (Mahowald et al., 2016), but this lexical-boost is not as long-lasting (Hartsuiker et al., 2008). Previous research on structural priming has heavily leaned on measures of syntactic choice, but a few studies have also measured initiation time (Smith & Wheeldon, 2001). The current study looks to differentiate effects of lexical and structural repetition benefits in initiation latency as a measure of potential fluency increases in sentence production. Results suggest that benefits in fluency are dependent on lexical and structural repetition.

Does Positional Level Structural Priming Depend on the Verb?

Complement coercion interpretation is typically operationalized in terms of verb paraphrases (she finished the book > she finished reading the book), reflecting the construal of a specific event triggered by the direct object (cf. Pustejovsky & Bouillon). However, when given the option, participants frequently, and selectively, refrain from providing verb paraphrases for naturally occurring sentences.

We present a corpus study of the pragmatic factors that affect the rate of refraining from giving a verb paraphrases, followed by two experiments. First, a fill-in-the-blank experiment shows that minimally different clause types containing complement coercion yield different rates of verb paraphrase. Second, we pair complement coercion sentences with possible verb paraphrases and obtain semantic (entailment) judgements from participants. Taken together, the results present a puzzle; participants’ likelihood to produce a verbal coercion interpretation is modulated by contextual factors, while their semantic comprehension of potential verb paraphrases for the same sentences is not.

Two measures for complement coercion interpretation: Interpretation vs production for complement coercion

Syntactic adaptation to short-term exposure has been documented with both single-trial priming and cumulative priming paradigms. These studies usually involve repeated exposure to the same structure (e.g. reduced relative clauses), and therefore it remains open whether people can track context-dependent regularities through short-term exposure. In the current study, we present a self-paced-reading experiment that investigates context-dependent syntactic adaptation by manipulating the relationship between the animacy feature of the subject NP (animate vs. inanimate) and the corresponding parse of a verb following a subject NP. We analyzed the results in terms of a log-linear model for context-dependent syntactic adaptation. The results suggest that comprehenders can track and adapt to cue-based distributional regularities, but only when the short-term regularities are consistent with the long-term ones existent in their native language.

Syntactic adaptation to short-term cue-based distributional regularities

Morphological syncretism occurs in languages when one morphological category ‘merges’ with another. Cross-linguistic research on the prevalence and types of syncretic patterns has revealed that some types of syncretism are more common than others. For example, syncretism in nominal morphology is more likely to occur in non-singular categories (Baerman et al., 2005). In two artificial language learning experiments, participants were exposed to words from a miniature language with suffix markings for gender (feminine and masculine) and number (singular, dual, and plural). Participants in Experiment 1 showed no evidence of a bias for syncretism in non-singular forms. However, participants in Experiment 2 showed a general bias to infer that a suffix that marked a novel category should be identical to a known form. This bias was strongest for non-singular items, in line with the cross-linguistic typology of syncretism. Implications for learnability and typology are discussed.

Learning Biases for Syncretic Morphological Systems

Previous studies have claimed that language structures tend to minimize the linear distance between syntactic heads and their dependents, a principle known as Dependency Length Minimization (DLM). These studies, however, have largely focused on written modality. In this study we examine the role of dependency length in acceptability ratings of English and Hindi, two typologically distinct languages, using audio stimuli. With double PP constructions as a test case, our results demonstrate no effect of DLM, suggesting the preference for shorter dependencies is different in acceptability and written texts. These findings are further supported with corpus analysis of a total of 10 treebanks for the two languages, which shows additional language-specific differences in the extent of DLM. We discuss the implications of our work and call for more careful consideration of linguistic and modality-specific diversity when it comes to processing-based claims about language typology.

Does One Size Fit all in Crosslinguistic Dependency Length Minimization?

Previous research has shown that when people engage in a joint task, they tend to minimize the total required effort. This is the case even when minimizing total effort increases the amount of effort individuals would need to expend. This tendency towards co-efficiency has previously been investigated in the motor domain. The aim of the current experiments was to investigate whether these findings extend to the sharing of mental effort. Five experiments were conducted using a multiple-object tracking task and two used a memory-based task. The first two experiments confirmed that, individually, participants prefer easier tasks. The following three experiments manipulated task difficulty for a participant and their ostensible partner and measured preferences for different difficulty combinations. Our findings provide support for egoistic effort distribution strategies, with participants mostly minimizing their own effort.

Premium content

Downloads

Next from CogSci 2022

Perceptual Similarity Affects Relational Judgements

Similar lecture

Scientific and Creative Analogies in Pretrained Language Models