Tibetan Relation Extraction for Medicine Entity of Multi-Granularity
-
Abstract
The complex sentence structures inherent in Tibetan medicine data make the relation extraction process highly sensitive to subtle semantic variations. To address this challenge, we propose the Tibetan relation extraction for multi-granular medicine entities. This method leverages a multi-head attention mechanism to extract entities at various granularities. A bilinear model is then employed to assign potential relationships to each entity pair. Additionally, a collaborative perception learning is introduced to improve training efficiency and accelerate convergence. Experimental results show that the proposed method significantly alleviates extraction sensitivity issues in Tibetan medicine, with robustness and adaptatbility further validated on the Chinese CMeIE-V2 dataset and the English NYT dataset.
-
-