GDMP: A Multi-Component Edge-Weighting Framework for Conceptual Shortest Paths in the Graph of Scientific Knowledge
Fuente:
Zenodo
Enregistré dans:
| Auteur principal: | |
|---|---|
| Format: | Recurso digital |
| Publié: |
Zenodo
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866901755294384128 |
|---|---|
| author | Raimov, Mikhail |
| author_facet | Raimov, Mikhail |
| contents | <p>The rapid expansion of scientific knowledge has created vast, interconnected graphs of concepts, theories, and publications. Traditional shortest-path algorithms, such as Dijkstra’s (1959), treat all edges as uniform and compute minimal hop counts or simple weighted distances. However, in the “Graph of Science”—a directed, multi-layered network spanning mathematics, physics, chemistry, and beyond—such metrics fail to capture conceptual distance, which must account for semantic similarity, abstraction levels, inter-layer transitions, uncertainty, influence, and temporal relevance.<br>We propose a novel multi-component edge-weighting framework that transforms classical single-objective shortest-path computation into a multi-objective optimization problem. Each edge is assigned a six-dimensional weight vector, enabling Pareto-optimal paths tailored to specific scientific inquiry modes (e.g., minimal-dependency, minimal-abstraction-jump, most-influential). The framework integrates seamlessly with recent breakthroughs in shortest-path algorithms, notably the deterministic O(m log^{2/3} n) directed single-source shortest path algorithm of Duan et al. (2025), which breaks the sorting barrier and scales to graphs with millions of nodes.<br>We demonstrate the framework’s applicability to (i) conceptual navigation across disciplines, (ii) automated discovery of interdisciplinary bridges, and (iii) large-scale citation and knowledge graphs derived from millions of scientific articles. In the latter, the model enables quantitative assessment of research novelty by measuring the conceptual cost of the shortest path from a new publication to the existing knowledge frontier.<br>This work lays the foundation for next-generation scientific navigation systems, AI-driven literature synthesis, and metrics for scientific progress.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18635312 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | GDMP: A Multi-Component Edge-Weighting Framework for Conceptual Shortest Paths in the Graph of Scientific Knowledge Raimov, Mikhail knowledge graphs shortest paths multi-objective optimization scientific discovery novelty detection graph of science <p>The rapid expansion of scientific knowledge has created vast, interconnected graphs of concepts, theories, and publications. Traditional shortest-path algorithms, such as Dijkstra’s (1959), treat all edges as uniform and compute minimal hop counts or simple weighted distances. However, in the “Graph of Science”—a directed, multi-layered network spanning mathematics, physics, chemistry, and beyond—such metrics fail to capture conceptual distance, which must account for semantic similarity, abstraction levels, inter-layer transitions, uncertainty, influence, and temporal relevance.<br>We propose a novel multi-component edge-weighting framework that transforms classical single-objective shortest-path computation into a multi-objective optimization problem. Each edge is assigned a six-dimensional weight vector, enabling Pareto-optimal paths tailored to specific scientific inquiry modes (e.g., minimal-dependency, minimal-abstraction-jump, most-influential). The framework integrates seamlessly with recent breakthroughs in shortest-path algorithms, notably the deterministic O(m log^{2/3} n) directed single-source shortest path algorithm of Duan et al. (2025), which breaks the sorting barrier and scales to graphs with millions of nodes.<br>We demonstrate the framework’s applicability to (i) conceptual navigation across disciplines, (ii) automated discovery of interdisciplinary bridges, and (iii) large-scale citation and knowledge graphs derived from millions of scientific articles. In the latter, the model enables quantitative assessment of research novelty by measuring the conceptual cost of the shortest path from a new publication to the existing knowledge frontier.<br>This work lays the foundation for next-generation scientific navigation systems, AI-driven literature synthesis, and metrics for scientific progress.</p> |
| title | GDMP: A Multi-Component Edge-Weighting Framework for Conceptual Shortest Paths in the Graph of Scientific Knowledge |
| topic | knowledge graphs shortest paths multi-objective optimization scientific discovery novelty detection graph of science |
| url | https://doi.org/10.5281/zenodo.18635312 |