Linked Data Ground Truth for Quantitative and Qualitative Evaluation of Explanations for Relational Graph Convolutional Network Link Prediction on Knowledge Graphs
Bibliographic record
Abstract
Relational Graph Convolutional Networks (RGCNs) identify relationships within a Knowledge Graph to learn real-valued embeddings for each node and edge. Recently, researchers have proposed explanation methods to interpret the predictions of these black-box models. However, comparisons across explanation methods for link prediction remains difficult, as there is neither a method nor dataset to compare explanations against. Furthermore, there exists no standard evaluation metric to identify when one explanation method is preferable to the other. In this paper, we leverage linked data to propose a method, including two datasets (Royalty-20k, and Royalty-30k), to benchmark explanation methods on the task of explainable link prediction using Graph Neural Networks. In particular, we rely on the Semantic Web to construct explanations, ensuring that each predictable triple has an associated set of triples providing a ground truth explanation. Additionally, we propose the use of a scoring metric for empirically evaluating explanation methods, allowing for a quantitative comparison. We benchmark these datasets on state-of-the-art link prediction explanation methods using the defined scoring metric, and quantify the different types of errors made with respect to both data and semantics.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.063 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.006 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.005 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".