Improving Knowledge Graph-based BERT Pre-Training Through Entity Hot Partition and Adapter Fusion
Bibliographic record
Abstract
Enhancing pre-training models based on knowledge graphs is an emerging technique used to handle knowledge-intensive tasks. In the latest research presented at EMNLP 2021, an approach to multi-subgraph BERT embedding using knowledge graph partitioning is developed, demonstrating the effectiveness of parallel multiple subgraphs-based BERT embeddings. However, this approach is task-agnostic and only consider topological features during the knowledge graph partitioning process, while in practical applications such as question answering, the graph partitioning can dynamically change as users' interaction, that may neglect entity popularity features and potentially affect the performance of pre-training models. In this paper, we propose an entity hot partition(EHP) approach to multi-subgraph BERT embedding. We improve the performance of the BERT model through integrating entity hot partitioning and adapter fusion. By calculating entity popularity based on knowledge graph triplets, we construct an entity popularity weight vector and introduce it into the classic METIS method to achieve popularity-sensitive graph partitioning. Finally, the partitioned graph results are fed into adapters to facilitate fine-tuning-based BERT pretraining. We conducted extensive experiments on three biomedical BERT models, namely SciBERT, BioBERT, and PubMedBERT, across six downstream tasks. The results demonstrate the effectiveness of our approach in achieving the performance improvements under simulated popularity scenarios.
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 | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.005 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 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".