We present the DCU team's system for the CASTLE Challenge at ACM Multimedia 2025, which explores video retrieval and question answering in egocentric, multi-user environments. Our system adapts techniques developed for lifelogging, particularly event-based semantic retrieval and QA pipelines, to the CASTLE dataset with minimal architectural changes. It combines vision-language embeddings, transcript-based retrieval, and person tracking to support both automatic and interactive search workflows. In the interactive track, we introduce a modular interface for narrative reconstruction and exploratory search. Qualitative results show that the system can generate plausible, evidence-based answers to complex multimodal queries. These findings suggest that lifelog retrieval systems offer a viable foundation for broader egocentric video analysis.

Our paper introduces a novel middleware designed to enhance the user experience of novice users of video event retrieval systems. The middleware simplifies the search process while maintaining advanced semantic information-seeking features. It achieves this by refining sentence representation, rewriting the original query based on user intent and dataset metadata knowledge, reducing ambiguities, and increasing relevance with domain-specific terms. Our approach allows users to provide simplified descriptions, while the system automates query refinement and parameterizes these for different search functions. This reduces the difficulty of navigating complex user interfaces and selecting from available functionalities. Additionally, the multilingual feature supports non-native English speakers by helping them describe their queries more effectively. In general, our system is suitable for people who are not experienced in advanced retrieval tasks, specifically focusing on its applicability in the special session Interactive Video Retrieval for Beginners (IVR4B).