SSDAU: Structured Semantic Data Augmentation for Joint Entity and Relation Extraction
A new method called Structured Semantic Data Augmentation (SSDAU) addresses weak generalization in Joint Entity and Relation Extraction (JERE) caused by low-quality training data. Existing data augmentation techniques often disrupt semantic structures and dependencies, failing to generate effective augmented data. SSDAU preserves semantic structure by segmenting text based on entity labels and using an encoder to capture entity features through context awareness. It performs entity semantic restructuring to create augmented data and fuses contextualized embeddings with traditional similarity scores to distinguish semantically similar entities. The method is detailed in a paper on arXiv (2605.23440).
Key facts
- SSDAU is a novel data augmentation method for JERE.
- It preserves semantic structure during augmentation.
- It segments text based on entity labels.
- It uses an encoder for context-aware entity features.
- It performs entity semantic restructuring.
- It fuses contextualized embeddings with similarity scores.
- The paper is on arXiv with ID 2605.23440.
- The method aims to improve model generalization.
Entities
Institutions
- arXiv