Four Papers Accepted at ESWC 2025

The following papers have been accepted:

• Simon Ott, Melisachew Wudage Chekol, Christian Meilicke, Heiner Stuckenschmidt: Training-free Score Calibration for Complex Query Decomposition
  Answering complex queries on incomplete knowledge graphs poses significant challenges, as models must infer their answers despite gaps in the available data. Previous research has addressed this problem by developing end-to-end architectures specifically designed for complex query answering. These models are difficult to interpret and require extensive data and computational resources for training. Alternatively, some approaches have focused on leveraging existing neural link predictors, which have been designed for simple queries, to handle complex queries. This approach reduces the amount of training examples needed and offers more transparent reasoning. However, the output scores of the neural link predictors may require calibration for effective interaction during the reasoning process and a special adaption function has to be learned to achieve this.  In this work, (i) we show that depending on the query type, standard normalization methods are equally as effective as learning an adaptation function. (ii) Furthermore, we replace the neural link predictor with a rule-based approach that does not require any score calibration. With such an approach we achieve new state-of-the-art results and increase the mean reciprocal ranks from 35.1% to 37.1% averaged across datasets and query types. (iii) We conduct comprehensive empirical analysis to support our claims.
• Antonis Klironomos, Baifan Zhou, Zhuoxun Zheng, Mohamed H. Gad-Elrab, Heiko Paulheim and Evgeny Kharlamov: ReaLitE: Enrichment of Relation Embeddings in Knowledge Graphs using Numeric Literals
  Most knowledge graph embedding (KGE) methods tailored for link prediction focus on the entities and relations in the graph, giving little attention to other literal values, which might encode important information. Therefore, some literal-aware KGE models attempt to either integrate numerical values into the embeddings of the entities or convert these numerics into entities during preprocessing, leading to information loss. Other methods concerned with creating relation-specific numerical features assume completeness of numerical data, which does not apply to real-world graphs. In this work, we propose ReaLitE, a novel relation-centric KGE model that dynamically aggregates and merges entities' numerical attributes with the embeddings of the connecting relations. ReaLitE is designed to complement existing conventional KGE methods while supporting multiple variations for numerical aggregations, including a learnable method.
• Rita T. Sousa and Heiko Paulheim: Multi-dataset and transfer learning using gene expression knowledge graphs for patient diagnosis
  Gene expression datasets offer insights into gene regulation mechanisms, biochemical pathways, and cellular functions. Additionally, comparing gene expression profiles between disease and control patients can deepen the understanding of disease pathology. Therefore, machine learning has been used to process gene expression data, with patient diagnosis emerging as one of the most popular applications. Although gene expression data can provide valuable insights, challenges arise because the number of patients in expression datasets is usually limited, and the data from different datasets with different gene expressions cannot be easily combined.
  This work proposes a novel methodology to address these challenges by integrating multiple gene expression datasets and domain-specific knowledge using knowledge graphs, a unique tool for biomedical data integration. Then, vector representations are produced using knowledge graph embedding techniques, which are used as inputs for a graph neural network and a multi-layer perceptron. We evaluated the efficacy of our methodology in three settings: single-dataset learning, multi-dataset learning, and transfer learning. The experimental results show that combining gene expression datasets and domain-specific knowledge improves patient diagnosis in all three settings.
• Martin Böckling, Heiko Paulheim and Sarah Detzler: GeoRDF2vec – Learning Location-Aware Entity Representations in Knowledge Graphs
  Many knowledge graphs contain a substantial amount of spatial entities, such as cities, buildings, or natural places. For many of those, exact geometries are stored in the knowledge graphs. When learning representations of entities, however, they are usually not taken into account by most approaches. In this paper, we present a variant of RDF2vec which utilizes geometries to learn location-aware embeddings of entities. Our approach expands for each reachable node different nodes by flooding over the graph starting from the geographic nodes. Based on the flooded graph a modified RDF2vec variant is applied which biases graph walks based on spatial weights. In evaluations with different benchmark datasets, we show that our approach both outperforms non-location-aware RDF2vec, as well as TransGeo.