Scholarlink

<p>Recently, ontology embeddings representing entities in a low-dimensional space have been proposed for ontology completion. However, the ontology embeddings for concept subsumption prediction do not address the difficulties of similar and isolated entities and fail to extract the global information of annotation axioms from an ontology. In this paper, we propose a self-matching training method and two ontology embedding models: <b>In</b>verted-index <b>M</b>atrix <b>E</b>mbedding (InME) and <b>Co</b>-occurrence <b>M</b>atrix <b>E</b>mbedding (CoME). The two embeddings capture the global and local information in annotation axioms by means of the occurring locations of each word in a set of axioms and the co-occurrences of words in each axiom. The self-matching training method increases the robustness of concept subsumption prediction for ontologies with sufficiently rich annotation axioms, especially when predicted superclasses are similar to subclasses or correspond to isolated entities. Our evaluation experiments show that the self-matching training method with InME outperforms the existing ontology embeddings for the GO and FoodOn ontologies and that the method with the concatenation of CoME and OWL2Vec* outperforms them for the HeLiS ontology.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A self-matching training method with annotation embedding models for ontology subsumption prediction

  • 期刊论文

  • 发表日期：

    2026年3月2日

查看全文
  • Yukihiro Shiraishi,
  • Ken Kaneiwa

摘要

Recently, ontology embeddings representing entities in a low-dimensional space have been proposed for ontology completion. However, the ontology embeddings for concept subsumption prediction do not address the difficulties of similar and isolated entities and fail to extract the global information of annotation axioms from an ontology. In this paper, we propose a self-matching training method and two ontology embedding models: Inverted-index Matrix Embedding (InME) and Co-occurrence Matrix Embedding (CoME). The two embeddings capture the global and local information in annotation axioms by means of the occurring locations of each word in a set of axioms and the co-occurrences of words in each axiom. The self-matching training method increases the robustness of concept subsumption prediction for ontologies with sufficiently rich annotation axioms, especially when predicted superclasses are similar to subclasses or correspond to isolated entities. Our evaluation experiments show that the self-matching training method with InME outperforms the existing ontology embeddings for the GO and FoodOn ontologies and that the method with the concatenation of CoME and OWL2Vec* outperforms them for the HeLiS ontology.