H2CE: Modeling Geo-Semantic Interactions for POI Reranking
What it is
H2CE is a heterogeneous two-stage cross-encoder architecture designed for real-time Point-of-Interest (POI) reranking in local search. It models interactions between lexical semantics, geospatial proximity, and numerical quality signals by fusing bucketized natural-language descriptors with exact scalar values via latent-space aggregation, followed by a two-stage scoring process that uses pointwise filtering and pairwise comparison for the top candidates.
The problem it solves
Traditional POI reranking methods struggle to effectively model the non-linear interactions between lexical semantics, geospatial proximity, and numerical quality signals (like ratings) under strict real-time latency constraints.
What you could use it for
- Reranking local search results to balance relevance, distance, and user ratings
- Improving the precision of map-based search engines for nearby businesses
- Optimizing search latency while maintaining high ranking quality for geo-semantic queries
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.