Multimodal learning for AI-generated POI descriptions in leisure walking routes

People’s experiences of an environment are shaped by features that meet their attention – sometimes called landmarks, or points of interest (POIs). These features are perceived, memorized and communicated according to individual interests. Computational systems, however, typically manage POI information only by type and name. Further descriptive information about these POIs is often missing or insufficient, particularly for lesser-known POIs. To address this limitation, we propose a multi-modal learning approach for an automated generation of contextualized POI descriptions. Focus on leisure walking, the approach aims to provide descriptions of POIs along walking routes that reflect walkers’ interests and support more informed route choices.