Search

Dependency trees reflect rich structural information, which can effectively guide models to understand text semantics and are widely used for relation extraction. However, existing dependency-based models suffer from noise in dependency trees, which distorts the modeling of contextual information, particularly when introducing dependency types, further exacerbating the propagation and accumulation of errors. In addition, rule-based pruning strategies used to eliminate noise in dependency trees may lose crucial information for capturing relational patterns, whereas unconstrained learning-based pruning strategies may introduce more noisy edges. These two aspects of dependency trees may cause models to produce semantic confusion, which is detrimental to relation extraction. In this study, we propose a Contextual Dependency-Aware Graph Convolutional Network (C-DAGCN) that works on complete dependency trees and perceives the importance of word dependencies from multiple angles without considering the noise in dependency trees. This study includes three parts. First, the sequence model captures semantic information for the input sentence and its words. We considered the appropriate outputs as word and sentence-level representations based on model characteristics. Second, the structure model enhances the description of the sentence structure and its guidance. We consider that position knowledge implies associations between words; therefore, we propose the Current Word Global Relative Position Knowledge (CW-GRPK) to globally enhance text structure modeling and reduce the models' overreliance on single noisy structure guidance. A dependency-aware module weighs heuristic syntactic knowledge and non-heuristic CW-GRPK to capture fine-grained word-level interactions that effectively distinguish words related to relational semantics. Subsequently, a dependency-guided module, based on the results of the dependency-aware layer, encodes a knowledge-enhanced word graph without losing global linguistic patterns. Third, we constructed high-level linguistic representations based on the results of the sequence and structure model in C-DAGCN as a relational representation and mapped it to the relational decision space for relational prediction. The experimental results demonstrate that our method outperforms nine strong baseline models on two benchmark datasets. Compared with the best baseline Attentive Graph Convolutional Networks (A-GCN) model (Tian et al., 2021), our proposed C-DAGCN model with 90.15 macro-average F1 and 79.52 micro-average F1 achieves 0.28 macro-average F1 and 0.47 micro-average F1 improvement on the SemEval2010-Task8 and ACE2005 datasets, respectively. 1 1 Our code is available at https://github.com/dymyyc/RE%5fC-DAGCN. The released code when run produces results which are the same as those reported in this paper. • A kind of position knowledge globally enhances text structure modeling. • A dependency-aware module captures fine-grained word-level interactions. • Our model constructs a high-level linguistic representation for relation extraction. • Our model outperforms nine strong baselines on the two benchmark datasets. [ABSTRACT FROM AUTHOR]

Graph Neural Networks (GNNs) have received considerable attention in recent years due to their unique ability to model both topologies and semantics in the graphs. In this paper, we explore the use of GNNs in learning low-dimensional node representations in Heterogeneous Information Networks (HINs), which retain rich semantic information across multiple types of nodes and relations. Existing methods for HINs generally focus on modeling heterogeneity at either the node level or the relation level, but not both. As a result, these methods often fall short of optimal performance. To address this issue, we propose a novel Semantic- and Relation-aware Heterogeneous Graph neural Network, dubbed SR-HGN, which jointly incorporates rich semantics preserved on nodes and relations. Our approach involves projecting the HINs into a low-dimensional vector space through two steps: node-level aggregation and type-level aggregation. The node-level aggregation employs an attention mechanism to create relation vectors by aggregating messages from neighborhoods connected via the same type of relation. The type-level aggregation leverages relation vectors to aggregate node representations. In particular, we introduce semantic-aware attention and relation-aware attention in the type-level aggregation to model the contributions of relation vectors, in order to simultaneously gain knowledge from node semantics and relational information. Unlike other approaches that rely on pre-defined meta-paths, our model can be readily applied to most real-world applications without requiring any domain knowledge. To validate the effectiveness of our proposed approach, we conducted extensive experiments on three public datasets. Experimental results demonstrate that the SR-HGN significantly outperforms state-of-the-art baselines on node classification and node clustering tasks. [ABSTRACT FROM AUTHOR]

Relational triple extraction is aimed at detecting entity pairs with relations from sentences, which is a key technology for large-scale knowledge graph construction. Recent studies focus on the overlapping triple problem, where multiple relational triples may have overlaps in a sentence. However, these methods disregard the bidirectionality of triple extraction, which may lead to extracting invalid triples. In addition, many relational triples are labeled in datasets of the triple extraction task, implying domain knowledge information of these datasets, but current methods rarely consider it. In this paper, we present a bidirectional relation-guided attention network with semantics and knowledge (BRASK) for relational triple extraction. BRASK is a bidirectional extraction framework that is based on multitask learning and contains forward and backward triple extraction tasks. Forward triple extraction and backward triple extraction are parallel and complementary, which can obtain predicted results with high confidence. We utilize semantic relations and knowledge relations as guidance in forward triple extraction and backward triple extraction, respectively, thus integrating general and domain knowledge into our model. In addition, we adopt an attention mechanism to learn fine-grained sentence representations for different relations. BRASK can solve the triple overlap problem and capture bidirectional dependencies between subjects and objects. Experimental results show that BRASK achieves new state-of-the-art results in two public datasets, which demonstrates its effectiveness. [ABSTRACT FROM AUTHOR]