Mobility-Embedded POIs: Learning What A Place Is and How It Is Used from Human Movement

TL;DR

ME-POIs addresses the gap in POI representations by learning context-independent embeddings that fuse static text semantics with longitudinal mobility signals. The model combines a Visit Sequence Encoder, global POI alignment via a contrastive objective, and a Multi-Scale Distribution Transfer mechanism to handle sparse POIs, plus direct anchor supervision and text alignment, all optimized jointly with a composite objective $L = L_{ME-POI} + \lambda_a L_{KL-anchor} + \lambda_s L_{KL-sparse} + \lambda_t L_{text-align}$. After pretraining, embeddings are frozen and task heads are trained for downstream map-enrichment tasks. Experiments on two large mobility datasets show consistent improvements over text-only and mobility-only baselines, including substantial gains on dynamic tasks like visit intent, and improved handling of long-tail POIs, demonstrating that POI function is a critical signal for generalizable POI representations. ME-POIs thus provides a versatile framework for enriching POI representations that can benefit automated map maintenance, urban analytics, and location-based services.

Abstract

Recent progress in geospatial foundation models highlights the importance of learning general-purpose representations for real-world locations, particularly points-of-interest (POIs) where human activity concentrates. Existing approaches, however, focus primarily on place identity derived from static textual metadata, or learn representations tied to trajectory context, which capture movement regularities rather than how places are actually used (i.e., POI's function). We argue that POI function is a missing but essential signal for general POI representations. We introduce Mobility-Embedded POIs (ME-POIs), a framework that augments POI embeddings derived, from language models with large-scale human mobility data to learn POI-centric, context-independent representations grounded in real-world usage. ME-POIs encodes individual visits as temporally contextualized embeddings and aligns them with learnable POI representations via contrastive learning to capture usage patterns across users and time. To address long-tail sparsity, we propose a novel mechanism that propagates temporal visit patterns from nearby, frequently visited POIs across multiple spatial scales. We evaluate ME-POIs on five newly proposed map enrichment tasks, testing its ability to capture both the identity and function of POIs. Across all tasks, augmenting text-based embeddings with ME-POIs consistently outperforms both text-only and mobility-only baselines. Notably, ME-POIs trained on mobility data alone can surpass text-only models on certain tasks, highlighting that POI function is a critical component of accurate and generalizable POI representations.

Figures (5)

• Figure 1: Illustration of ME-POIs.ME-POIs augment static text-based POI representations with mobility-derived signals, to learn POI embeddings that capture their identity and function.
• Figure 2: Overview of ME-POIs pretraining. The framework includes: (i) a transformer-based visit sequence encoder, (ii) contrastive alignment of contextualized visits ($h$) with global POI embeddings ($z_{p}^{\text{ME}}$) to capture usage patterns, (iii) multi-scale distribution transfer to propagate temporal visit information to under-visited POIs ($p_s$), (iv) direct supervision on anchor POIs ($p_a$) to regularize embeddings via visit distribution prediction, and (v) an auxiliary text-alignment objective to ground POI embeddings ($z_{p}^{\text{ME}}$) in textual semantics ($z_{p}^{\text{text}}$).
• Figure 3: Impact of $\mathcal{L}_{\text{KL-anchor}}$ and $\mathcal{L}_{\text{KL-sparse}}$ on sparse and anchor POIs on open hours in Houston.
• Figure 4: Predictions on sparse and anchor POIs across models on open hours in Houston.
• Figure 5: Example prompt for Taco Man POI in Los Angeles.