CoChat

Emergence of analogy from relation learning

Open access
Lu, Hongjing, Wu, Ying Nian, Holyoak, Keith J. · 2019 · Proceedings of the National Academy of Sciences

The DOI matches its Crossref record; 41 of the 41 references it cites were found in a registry or an index.

Abstract

Significance: The ability to learn and make inferences based on relations is central to intelligence, underlying the distinctively human ability to reason by analogy across dissimilar situations. We have developed a computational model demonstrating that abstract relations, such as synonymy and antonymy, can be learned efficiently from semantic feature vectors for individual words and can be used to solve simple verbal analogy problems with close to human-level accuracy. The approach illustrates the potential synergy between deep learning from “big data” and supervised learning from “small data.” Core properties of high-level intelligence can emerge from relatively simple computations coupled with rich semantics. The model illustrates how operations on nonrelational inputs can give rise to protosymbolic relational representations.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1073/pnas.1814779116

Cites

41 of 41 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1. Illustration of the BART model for learning a semantic relation, R, from feature vectors for word pairs. Colors indicate features and weights associated, respectively, with the first word (red) and the second word (blue). Note that semantic roles (parts of networks based on first word and second word, respectively) are distinctly separated in the distribution of relation weights. The feature alignment problem is addressed by augmenting the raw feature vectors (layer 1 at bottom) by concatenating features ranked by difference (layer 2), followed by feature selection (layer 3), and learning of relational weight distributions using the variational Bayesian method with a contrast-based prior.
Fig. 1. Illustration of the BART model for learning a semantic relation, R, from feature vectors for word pairs. Colors indicate features and weights associated, respectively, with the first word (red) and the second word (blue). Note that semantic roles (parts of networks based on first word and second word, respectively) are distinctly separated in the distribution of relation weights. The feature alignment problem is addressed by augmenting the raw feature vectors (layer 1 at bottom) by concatenating features ranked by difference (layer 2), followed by feature selection (layer 3), and learning of relational weight distributions using the variational Bayesian method with a contrast-based prior.CC BY-NC-ND · PubMed Central
Fig. 2. Predictions of relation typicality. Correlations between human typicality ratings and model predictions for 10 types of abstract relations after training with 20 positive examples for each relation by BART and for the baseline Word2vec model.
Fig. 2. Predictions of relation typicality. Correlations between human typicality ratings and model predictions for 10 types of abstract relations after training with 20 positive examples for each relation by BART and for the baseline Word2vec model.CC BY-NC-ND · PubMed Central
Fig. 3. Illustration of relation vectors, in which each element indicates how likely a word pair is to instantiate a relation. High probabilities are represented in yellow; low probabilities, in black. Relation vectors are shown for four positive examples of the specific contradictory relation, ordered by typicality with respect to that relation. As the typicality of the word pairs with respect to the contradictory relation declines, the relation vector becomes increasingly distributed across a broader range of relation types.
Fig. 3. Illustration of relation vectors, in which each element indicates how likely a word pair is to instantiate a relation. High probabilities are represented in yellow; low probabilities, in black. Relation vectors are shown for four positive examples of the specific contradictory relation, ordered by typicality with respect to that relation. As the typicality of the word pairs with respect to the contradictory relation declines, the relation vector becomes increasingly distributed across a broader range of relation types.CC BY-NC-ND · PubMed Central
Fig. 4. Proportion of correct solutions to UCLA VAT problems achieved by BART and by Word2vec for each of four relation types. An accuracy level of 0.50 represents chance performance. Error bars for human performance indicate 95% confidence intervals.
Fig. 4. Proportion of correct solutions to UCLA VAT problems achieved by BART and by Word2vec for each of four relation types. An accuracy level of 0.50 represents chance performance. Error bars for human performance indicate 95% confidence intervals.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 41 references

Research works printed but not linked: 2

  1. AA Petrov Associative Memory-Based Reasoning: A Computational Model of Analogy-Making in a Decentralized Multi-Agent Cognitive Architecture (Lambert Academic, Saarbrücken, Germany, 2013).
  2. 10.1037/0096-3445.104.3.192 · doi:10.1037/0096-3445.104.3.192 · Paper page

Cited by

See all 84

Metadata

Authors
Lu, H., Wu, Y. N., & Holyoak, K. J.
Affiliations
University of California, Los Angeles (US)
Corresponding author
Lu, Hongjing
Venue
Proceedings of the National Academy of Sciences
Published
2019-02-15
Volume
116
Issue
10
Pages
4176-4181
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
OpenAlex
W2913142708
Cited by
70
References
38
Citations by year
2026: 1 · 2025: 9 · 2024: 7 · 2023: 12 · 2022: 10 · 2021: 10 · 2020: 14 · 2019: 7
FWCI
4.68
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
Funders
  • Google (Google Faculty Research Award)
  • NSF | Directorate for Social, Behavioral and Economic Sciences (BCS-1827374)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Categorization, perception, and language
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
Keywords
analogy, relational learning, feature alignment, relational similarity, verbal analogies, antonymy, category membership, bootstrapping, computational model of language acquisition, middle childhood, typicality judgments

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Emergence of analogy from relation learning. https://app.cochat.ai/papers/10.1073/pnas.1814779116
The paper
Lu, H., Wu, Y. N., & Holyoak, K. J. (2019). Emergence of analogy from relation learning. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.1814779116

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.