Representing the disciplinary structure of physics
A comparative evaluation of graph and text embedding methods
Bibliographic Data
| ID | 21255825 |
|---|---|
| Authors | Isabel Constantino (0000-0002-4988-7146, Indiana University Bloomington), Sadamori Kojaku (0000-0002-9414-6814, Binghamton University), Santo Fortunato (0000-0002-9039-4730, Indiana University Bloomington), Yong-Yeol Ahn (0000-0002-4352-4301, Indiana University Bloomington) |
| Year | 2025 |
| Volume | 6 |
| Pages | 263-280 |
| Publication date | 2025-03-12 |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Quantitative Science Studies (JOURNAL) |
| Journal identifiers | ISSN: 2641-3337 • E-ISSN: 2641-3337 |
| Publisher | MIT Press (PUBLISHER • US) |
| DOI | 10.1162/qss_a_00349 |
| OpenAlex | W4406721862 |
| Language | EN |
| Citations received | 1 |
| References cited | 48 |
Recent advances in machine learning offer new ways to represent and study scholarly works and the space of knowledge. Graph and text embeddings provide a convenient vector representation of scholarly works based on citations and text. Yet, it is unclear whether their representations are consistent or provide different views of the structure of science. Here, we compare graph and text embedding by testing their ability to capture the hierarchical structure of the Physics and Astronomy Classification Scheme (PACS) of papers published by the American Physical Society (APS). We also provide a qualitative comparison of the overall structure of the graph and text embeddings for reference. We find that neural-network-based methods outperform traditional methods, and graph embedding methods node2vec and residual2vec are better than other methods at capturing the PACS structure. Our results call for further investigations into how different contexts of scientific papers are captured by different methods, and how we can combine and leverage such information in an interpretable manner
Discipline · Embedding · Graph · Physics · Social science · Sociology · Advanced Text Analysis Techniques · Computer Science · Topic Modeling · Artificial Intelligence · Theoretical Computer Science
SciBert
The link‐prediction problem for social networks
Finding scientific topics
Representation Learning
Large teams develop and small teams disrupt science and technology
Sentence-Bert
Recombinant Growth
Mapping the backbone of science
Bias against novelty in science
DeepWalk
The Structure of Scientific Literatures I
Atypical Combinations and Scientific Impact
Recombinant Uncertainty in Technological Search
Visualizing Bibliometric Networks
Bibliographic coupling between scientific papers
Individual Comparisons by Ranking Methods
Node2vec
Co‐citation in the scientific literature
On Information and Sufficiency
Probabilistic topic models
Semantic and relational spaces in science of science
A comparison of large-scale science models based on textual, direct citation and hybrid relatedness
A principled methodology for comparing relatedness measures for clustering publications
Friends and neighbors on the Web
| Citations per year | 1 |
|---|---|
| Citation span | 2025 - 2025 (1) |
| Highly cited | No |