SCRE: special cargo relation extraction using representation learning

Abstract The airfreight industry of shipping goods with special handling needs, also known as cargo, often deals non-transparent data and outdated technology, resulting in significant inefficiency. A cargo ontology is a means extracting, structuring, storing domain knowledge representing the concepts relationships that can be processed by computers. This used base semantic retrieval many artificial intelligence applications, such planning for shipments. Domain information extraction an essential task implementing maintaining ontology. However, absence makes instantiating challenging. We propose relation representation learning approach based on hierarchical attention-based multi-task model leverage it domain. proposed architecture applied identifying categorizing samples various types trained domain-specific documents number tasks vary from lightweight bottom layers to heavyweight top setting. Therefore, conveys complementary input features learns rich representation. train relies only entity-linked corpus shipment These two models are then employed supervised multi-class classifier called Special Cargo Relation Extractor (SCRE). results experiments show represent complex efficiently.