GeoAI with OSM: Harmonizing POI Categories + Measuring Tag Fragmentation

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Bachelorarbeit / Semesterarbeit / Masterarbeit /

Inhalt

experimentell / theoretisch / konstruktiv /  

Beschreibung

OSM POI tags are open, inconsistent, and region-dependent: the same place type is tagged
differently across cities/countries, and different place types can share similar tags. This creates
noisy labels for GeoAI pipelines (e.g., neighborhood embeddings, land-use inference, urban
function classification) and reduces comparability between study areas.
The goal of this thesis is to harmonize heterogeneous OSM POI tags into a clean set of ca-
nonical POI classes (“POI bundles”) and to quantify how fragmented the original tagging is
across regions.
Possible methods include (i) rule/ontology-based mapping (tag normalization + synonym ta-
bles), (ii) geographical context-aware ML using multi-scale spatial neighborhoods around each
POI (e.g., ring-based context features) to reclassify into canonical classes, and (iii) semi-su-
pervised clustering/embeddings with a small labeled set to calibrate clusters to the taxonomy.
Evaluation focuses on mapping coverage, cross-region robustness, and fragmentation metrics
per class and region.

Workpackages

• Literature review on OSM POI harmonization and POI classification
• Extract and preprocess POIs for multiple study areas; normalize tags
• Define a canonical POI taxonomy (flat or hierarchical)
• Implement a harmonization pipeline
• Evaluate harmonization (coverage, cross-region generalization)
• Quantify and visualize tag fragmentation (per class, per region)
• Discuss results and deliver a reusable schema + dataset export

Voraussetzungen

Programming skills in Python
Structured and independent working style

Tags

FTM Studienarbeit, FTM SM, FTM Zacher, FTM Informatik

Möglicher Beginn

sofort

Kontakt

Till Zacher, M.Sc.
Tel.: +49 89 289 15351
till.zachertum.de

Ausschreibung