Abstract: Traditional knowledge graph embedding methods were mainly focused on link prediction for simple triples, and significant limitations in handling conjunctive queries containing multiple unknown variables were had by their modeling paradigm of “head entity-relation-tail entity”. To address the above issues, a complex query question-answering model integrating Bidirectional Sequence Embeddings (BSE) was proposed. First, a query encoder based on a bidirectional Transformer architecture was constructed to convert the query structure into a serialized representation. Second, positional encoding was utilized to preserve graph structure information. Third, the deep semantic associations among all elements in the query graph were dynamically modeled through the Additive Attention Mechanism (AAM). Finally, global information interaction across nodes was realized, and the shortcomings of traditional methods in modeling long-distance dependencies were effectively addressed. Experiments are conducted on different benchmark datasets to verify the performance advantages of the BSE model. Specifically, on the WN18RR-PATHS dataset, compared with GQE-DistMult-MP, the BSE model achieves a 48.99% improvement in the MRR (Mean Reciprocal Rank) metric. In addition, on the EduKG education dataset, the BSE model outperforms GQE-Bilinear by a 6.08% increase in the AUC (Area Under the Curve) metric. To sum up, the proposed model can be applied to query-based question answering in different fields, and has high scalability and application value.

Key words: Knowledge Graph (KG), bidirectional sequence, semantic association, long-distance dependency, GQE-DistMult-MP 

 

摘要： 传统知识图谱嵌入方法主要聚焦于简单三元组的链接预测，它的“头实体-关系-尾实体”的建模范式在处理包含多个未知变量的合取查询时存在显著局限性。针对上述问题，提出融合双向序列嵌入(BSE)的复杂查询问答模型。首先，基于双向Transformer架构构建查询编码器，将查询结构转换为序列化表示；其次，利用位置编码保留图结构信息；再次，通过加法注意力机制(AAM)动态建模查询图中所有元素的深层语义关联；最后，实现跨节点的全局信息交互，以解决传统方法在长距离依赖建模方面的缺陷。在不同基准数据集上进行实验，验证BSE模型的性能优势。其中，在WN18RR-PATHS数据集上，BSE与GQE-DistMult-MP相比，平均倒数排名(MRR)指标提高了48.99%；在EduKG教育数据集上，BSE与GQE-Bilinear相比，曲线下面积(AUC)指标提高了6.08%。综上所述，所提模型可用于不同领域的查询问答，具有较高扩展性与应用价值。

关键词: 知识图谱, 双向序列, 语义关联, 长距离依赖, GQE-DistMult-MP