H2CE: Modeling Geo-Semantic Interactions for POI Reranking with Heterogeneous Two-Stage Cross-Encoders

2026-10-08

AI Summary

This study addresses the challenges of multimodal feature fusion and real-time latency constraints in point-of-interest (POI) reranking for local search by proposing a heterogeneous two-stage cross-encoder. The method innovatively introduces bucketized numerical text to enhance semantic attention and integrates scalar MLP embeddings to achieve nonlinear interactions through latent-space aggregation. Furthermore, it adopts a two-stage architecture that reduces pairwise comparison complexity from quadratic to linear, effectively balancing accuracy and efficiency. Experimental results demonstrate that the proposed model achieves an NDCG@5 of 67.48% on the test set, outperforming XGBoost by 22.82% and significantly surpassing zero-shot large language model-based rerankers.

Abstract

Point-of-Interest (POI) reranking in local search must model query-conditioned tradeoffs among lexical semantics, geospatial proximity, and numerical quality signals such as rating and review count, while remaining practical under real-time serving constraints. A close POI may only partially satisfy the query intent, while a farther one may offer stronger semantic and quality evidence. We present H2CE, a Heterogeneous Two-stage Cross-Encoder for latency-bounded POI reranking. H2CE represents numerical attributes in two complementary ways: bucketized natural-language descriptors are inserted into the cross-encoder input to support semantic--numeric attention, while exact scalar values are processed by dedicated MLPs to preserve magnitude information. The resulting semantic and numerical embeddings are fused through latent-space aggregation, enabling nonlinear interactions beyond scalar weighted sums. H2CE then applies a two-stage architecture: Stage 1 scores all candidates pointwise for scalable filtering, and Stage 2 performs head-to-head pairwise comparison among the top-$K$ candidates with Copeland aggregation, making fine-grained relative tradeoffs explicit while reducing pairwise cost from O(N^2) to O(N+K(K-1)). On a 5,743-query local search test set, H2CE achieves 67.48% NDCG@5, improving over XGBoost LTR by +22.82% absolute and over a zero-shot LLM reranker by +35.89%. The pairwise stage adds +1.98% NDCG@5 over the pointwise model alone. Ablations confirm the value of numerical features, latent aggregation, top-K pairwise reranking, and aligned training.

Problem

• POI reranking
• local search
• geo-semantic interactions
• numerical quality signals
• real-time serving constraints

Innovation

• Cross-Encoder
• POI Reranking
• Two-Stage Architecture
• Geo-Semantic Interactions
• Latent-Space Aggregation