Yan Zhou

• Lyu, Qianhang; Skjæveland, Martin G; Zhou, Yan; Rao, Yunqing; Waaler, Arild Torolv Søetorp & Zhou, Baifan (2025). IMF-PM: Integrating System Information Modelling and Project Management for Industrial Digitalisation. In Pang, Zhibo & Vyarkin, Valeriy (Ed.), 2025 IEEE 23rd International Conference on Industrial Informatics (INDIN). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331511227. doi: 10.1109/INDIN64977.2025.11279421. Full text in Research Archive Show summary

  In the context of Industry 4.0 and digital transformation, project management (PM) and system information modelling (SIM) are two key approaches for managing complex, crossdisciplinary industrial systems. However, these approaches are often applied separately, and limited research has explored their integration. To bridge this gap, this paper proposes the IMF-PM methodology, integrating the Information Modelling Framework (IMF) with the PMBOK project management standard, to enable structured, and model-based project management. IMF-PM features an engineer-friendly, multi-aspect modelling language that captures product, function, and attribute information across disciplines. Its simple yet expressive elements enable engineers from different disciplines to model systems within a unified framework, promoting interoperability and collaboration. Additionally, the method supports semantic serialisation into RDF triples, enhancing digitalisation and supporting the creation of digital twins of real-world systems. The approach is demonstrated through a real-world use case on designing a high-temperature gas cooling system and compared against alternative methods to highlight its practical advantages.

• Zhou, Yan; Zhou, Baifan & Yu, Ingrid Chieh (2025). Towards Change-Instructed Action and Explanation Generation for Visual Language Action in Autonomous Driving. In Pang, Zhibo & Vyarkin, Valeriy (Ed.), 2025 IEEE 23rd International Conference on Industrial Informatics (INDIN). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331511227. doi: 10.1109/indin64977.2025.11279112. Full text in Research Archive
• Zhou, Yan; Zhou, Baifan; Lyu, Qianhang; Waaler, Arild Torolv Søetorp & Yu, Ingrid Chieh (2025). User-Centric Question Answering with Explanation for Industrial System Information Modelling. Frontiers in Artificial Intelligence and Applications. ISSN 0922-6389. 413, p. 5392–5399. doi: 10.3233/faia251478. Full text in Research Archive Show summary

  AI holds great promise for industrial applications, but its complexity and opacity can hinder trust and limit adoption, particularly in high-stakes sectors, where transparency and accountability are essential. Question Answering with Explanation (QAE) helps address this challenge by enabling natural interaction with AI and providing justifications that allow users to assess, verify, and better understand AI-generated answers, thus improving perceived credibility. However, despite its clear benefits, QAE remains underexplored in industrial contexts. To address this gap, we employ a user-centric framework for QAE, and study it in system information modelling (SIM), a realistic and valuable industrial use case spanning disciplines, lifecycle stages, and downstream engineering tasks. Our framework integrates retrieval-augmented generation (RAG), with instruction, chain-of-thought, and few-shot prompting to guide LLMs in producing reasoning-rich explanations, and incorporates both automatic and user-perspective human evaluation. Extensive experiments on PDF and RDF datasets covering five key SIM topics demonstrate near-human-level performance. Case studies further illustrate our strengths in delivering strong argumentation, contextual evidence, and complex reasoning chains.

• Zhou, Yan; Zhou, Baifan & Yu, Ingrid Chieh (2025). NMN-BART:Generating Natural Language Explanations for Visual Question Answering. CEUR Workshop Proceedings. 3975, p. 38–48. doi: https:/ceur-ws.org/Vol-3975/paper4.pdf. Full text in Research Archive Show summary

  Visual Question Answering (VQA) is a challenging task that requires reasoning over both visual and textual information. Recently, there has been growing interest in enhancing VQA with Natural Language Explanations (NLEs) to improve transparency and trust. While existing methods leverage powerful language models for explanation generation, many score high on lexical-level text similarity rather than capturing the underlying reasoning process. In this work, we propose NMN-BART, a novel architecture that combines Neural Module Networks (NMNs) with the pretrained BART language model, using cross-modal fusion to bridge visual semantics and textual reasoning. We evaluate NMN-BART on the VQA-X dataset, where it significantly outperforms baselines on semantic-based metrics, despite lower scores on lexical similarity metrics. This suggests that our method excels in capturing the meaningful content of the explanations, rather than matching the references in wording. The case study with human evaluation further verifies our finding that our method produces semantically rich and persuasive explanations.

• Zhou, Yan; Zhou, Baifan; Li, Huajian; Lyu, Qianhang; Qu, Yuanwei & Waaler, Arild Torolv Søetorp [Show all 7 contributors for this article] (2025). Dataset for Industrial Question Answering with Explanation and Scalable Ensemble Generation. In Long, Guodong; Blumestein, Michale & Chang, Yi (Ed.), WWW '25: Companion Proceedings of the ACM on Web Conference 2025. Association for Computing Machinery (ACM). ISSN 9798400713316. p. 825–828. doi: 10.1145/3701716.3715310. Full text in Research Archive Show summary

  The digital and green transition under Industry 4.0 has accelerated the adoption of AI in industries such as manufacturing, energy, and mining. Question Answering with Explanation (QAE), as a way of human interaction with AI, is crucial for enhancing transparency and trust in high-stakes industrial applications. However, industrial QAE remains underexplored due to the lack of publicly available, high-quality datasets, hindered by the need for expert effort and corporate restrictions. To this end, we introduce PANDAX ( https://doi.org/10.5281/zenodo.14510798 ), the first open-source industrial QAE dataset, and SEG, a scalable method for generating high-quality QAE datasets using LLMs. PANDAX focuses on three key topics of industrial system information: partonomy, functionality, and parameters, across critical domains such as green technology and cooling systems. SEG ensures scalability and quality through ensemble generation, majority voting, expert ranking, etc. The human evaluation validates PANDAX's high quality, positioning it as a valuable resource for advancing QAE techniques, benchmarking language technologies, and supporting research in explainable AI for industrial systems.

View all works in NVA

• Qu, Yuanwei; Zheng, Zhuoxun; Zhou, Baifan; Zhou, Yan; Santos, Nicolau & Savkovic, Ognjen [Show all 8 contributors for this article] (2023). Towards Data Integrity Verification for More Sustainable Petroleum Industry. Full text in Research Archive
• Zhou, Yan (2026). Neuro-Symbolic Methods for Natural Language Explanations in Question Answering. Universitetet i Oslo. Full text in Research Archive Show summary

  The deployment of AI in high-stakes sectors requires trustworthiness and accountability. This thesis focuses on generating faithful, grounded and adaptable natural language explanations for question answering tasks using neuro-symbolic methods, enabling users to better understand, inspect, and validate the system’s decisions. Neuro-symbolic methods are architectures that combine the flexibility and learning capacity of neural components with the transparency and reasoning precision of symbolic components. This thesis shows how these methods can improve the consistency between the explanations and the system's reasoning process. It also investigates how different forms of domain knowledge can be represented through neural or symbolic techniques, and how they can be effectively incorporated to enhance the quality and usefulness of explanations. Depending on domain constraints and requirements, different model adaptation methods are employed, and new evaluation frameworks focusing on user trust are proposed beyond existing automated metrics.

View all works in NVA