Reminiscens, an AI-Powered Exploratory Search Engine for Memory-Triggering Artworks

Reminiscens is an AI-powered, multilingual exploratory search engine developed through an academic-museum collaboration to support a range of use cases such as memory-triggering activities, curatorial research, and public engagement. The system combines trustworthy metadata filters, semantic prompts (including negative cues), lightweight relevance feedback, and collection-building tools. To address sparse and uneven multilingual metadata, we generate synthetic captions that are merged with curated iconographic descriptions, and adapt three CLIP variants for use in cultural heritage retrieval. In parallel, we introduce a curator-in-the-loop suggestion assistant that uses the same embedding space to propose controlled-vocabulary iconographic terms, supporting metadata enrichment while keeping expert judgement central.