Google Research, in collaboration with the University of Southern California, has developed ME-POIs, a novel framework that enhances text-based place embeddings by incorporating aggregate human movement data. This approach aims to capture how a location is used, beyond its textual description, by encoding individual visits into contextualized vectors and aligning them with a learnable prototype for each Point of Interest (POI). Experiments on mobility data from Los Angeles and Houston demonstrated significant improvements across various map-enrichment tasks, with ME-POIs outperforming existing text encoders and even a mobility-only variant surpassing Gemini embeddings in price-level classification. AI
IMPACT Enhances location-based AI applications by providing richer, context-aware place representations.
RANK_REASON The cluster describes a research paper introducing a new framework for place embeddings. [lever_c_demoted from research: ic=1 ai=1.0]
- Gemini
- GeoLLM
- Google Research
- Houston
- Los Angeles
- ME-POIs
- NVIDIA Tesla V100 16GB
- Space2Vec
- Time2Vec: Learning a Vector Representation of Time
- Transformer++
- University of Southern California
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
