Shenghui Wang

Elena Malnatsky, Shenghui Wang, Koen V. Hindriks et al.

Dialogic learning fosters motivation and deeper understanding in education through purposeful and structured dialogues. Foundational models offer a transformative potential for child-robot interactions, enabling the design of personalized, engaging, and scalable interactions. However, their integration into educational contexts presents challenges in terms of ensuring age-appropriate and safe content and alignment with pedagogical goals. We introduce a hybrid approach to designing personalized educational dialogues in child-robot interactions. By combining rule-based systems with LLMs for selective offline content generation and human validation, the framework ensures educational quality and developmental appropriateness. We illustrate this approach through a project aimed at enhancing reading motivation, in which a robot facilitated book-related dialogues.

Floris Vossebeld, Shenghui Wang

Generating complex, logically-sound SPARQL queries for multi-hop questions remains a critical bottleneck for Knowledge Graph Question Answering, as the brittle nature of one-shot generation by Large Language Models (LLMs) hinders reliable interaction with structured data. Current methods lack the adaptive policies needed to dynamically debug queries based on real-time execution feedback. This paper introduces a novel agentic framework where an LLM learns a resilient policy for the sequential process of iterative SPARQL construction. We show that a compact 3B-parameter model, trained exclusively via outcome-driven Reinforcement Learning (GRPO) without supervised fine-tuning, can learn effective policies for this task, discovering how to systematically recover from execution errors and refine its queries toward a correct answer. On a curated, executable single-answer subset of LC-QuAD 2.0, our agent achieves 49.7\% accuracy post-entity-linking, a significant 17.5 percentage point improvement over the strongest iterative zero-shot baseline. Further analysis reveals that while the agent's capability is driven by RL, its performance is enhanced by an explicit deliberative reasoning step that acts as a cognitive scaffold to improve policy precision. This work presents a generalizable blueprint for teaching agents to master formal, symbolic tools through interaction, bridging the gap between probabilistic LLMs and the structured world of Knowledge Graphs.

Linfeng Li, Peng Wang, Yao Wang et al.

This paper proposes an algorithm named as PrTransH to learn embedding vectors from real world EMR data based medical knowledge. The unique challenge in embedding medical knowledge graph from real world EMR data is that the uncertainty of knowledge triplets blurs the border between "correct triplet" and "wrong triplet", changing the fundamental assumption of many existing algorithms. To address the challenge, some enhancements are made to existing TransH algorithm, including: 1) involve probability of medical knowledge triplet into training objective; 2) replace the margin-based ranking loss with unified loss calculation considering both valid and corrupted triplets; 3) augment training data set with medical background knowledge. Verifications on real world EMR data based medical knowledge graph prove that PrTransH outperforms TransH in link prediction task. To the best of our survey, this paper is the first one to learn and verify knowledge embedding on probabilistic knowledge graphs.

Rob Koopman, Shenghui Wang

After a clustering solution is generated automatically, labelling these clusters becomes important to help understanding the results. In this paper, we propose to use a Mutual Information based method to label clusters of journal articles. Topical terms which have the highest Normalised Mutual Information (NMI) with a certain cluster are selected to be the labels of the cluster. Discussion of the labelling technique with a domain expert was used as a check that the labels are discriminating not only lexical-wise but also semantically. Based on a common set of topical terms, we also propose to generate lexical fingerprints as a representation of individual clusters. Eventually, we visualise and compare these fingerprints of different clusters from either one clustering solution or different ones.

Judit Bar-Ilan, Rob Koopman, Shenghui Wang et al.

This panel brings together experts in bibliometrics and information retrieval to discuss how each of these two important areas of information science can help to inform the research of the other. There is a growing body of literature that capitalizes on the synergies created by combining methodological approaches of each to solve research problems and practical issues related to how information is created, stored, organized, retrieved and used. The session will begin with an overview of the common threads that exist between IR and metrics, followed by a summary of findings from the BIR workshops and examples of research projects that combine aspects of each area to benefit IR or metrics research areas, including search results ranking, semantic indexing and visualization. The panel will conclude with an engaging discussion with the audience to identify future areas of research and collaboration.

Rob Koopman, Shenghui Wang, Andrea Scharnhorst

This paper builds on an innovative Information Retrieval tool, Ariadne. The tool has been developed as an interactive network visualization and browsing tool for large-scale bibliographic databases. It basically allows to gain insights into a topic by contextualizing a search query (Koopman et al., 2015). In this paper, we apply the Ariadne tool to a far smaller dataset of 111,616 documents in astronomy and astrophysics. Labeled as the Berlin dataset, this data have been used by several research teams to apply and later compare different clustering algorithms. The quest for this team effort is how to delineate topics. This paper contributes to this challenge in two different ways. First, we produce one of the different cluster solution and second, we use Ariadne (the method behind it, and the interface - called LittleAriadne) to display cluster solutions of the different group members. By providing a tool that allows the visual inspection of the similarity of article clusters produced by different algorithms, we present a complementary approach to other possible means of comparison. More particular, we discuss how we can - with LittleAriadne - browse through the network of topical terms, authors, journals and cluster solutions in the Berlin dataset and compare cluster solutions as well as see their context.