Multi-Objective Ranking to Boost Navigational Suggestions in eCommerce AutoComplete (WWW 2023)
This entry indexes the WWW 2023 paper "Multi-Objective Ranking to Boost Navigational Suggestions in eCommerce AutoComplete." The original PDF is available here:
Key points
- Problem domain: Ranking for eCommerce autocomplete, where navigational suggestions (queries that lead users directly to products or categories) must be ranked while balancing multiple objectives (user engagement, business metrics, intent coverage).
- Venue: WWW 2023; shared through the Seller Sessions resource library, targeting the intersection of academic IR research and industrial search practice.
- Positioning: The entry situates the paper within the modern learning-to-rank stack — candidate generation, ranking/reranking, and generation — and notes how LLM-era systems now treat retrieval strategy and inference budget as learnable decisions.
- Typical evaluation metrics: nDCG@10, MRR, Recall@k, Hit@k, plus online signals (CTR, satisfaction, task success) and system constraints (latency, token/compute cost).
- Common baselines: BM25, dense retrieval, cross-encoder reranking, retrieval-free LLMs, and commercial search APIs.
- Engineering checklist: PII handling and embedding version management, p99 latency budgets with cascading and early stopping, interleaved online experiments to validate offline gains, source whitelisting and output filtering for safety, and model routing/distillation for cost control.
- Deep Learning to Rank in Industrial Search Engines (ACM)
- Multi-Objective Recommendation in the Era of Generative AI: A Survey (arXiv 2506.16893)
- A Generative Re-ranking Model for List-level Multi-objective Optimization (arXiv 2505.07197)
- A Thorough Comparison of Cross-Encoders and LLMs for Reranking SPLADE (arXiv 2403.10407)
- Accelerating Listwise Reranking: Reproducing and Enhancing FIRST (SIGIR)
- Adaptive Neural Ranking Framework: Toward Maximized Business Goal (CIKM, 10.1145/3589334.3645605)
Context for practitioners
The surrounding analysis in the source post highlights considerations relevant to anyone implementing multi-objective suggestion ranking:
Related entries cross-referenced in the source post
Note on fidelity
This post is primarily a curated metadata entry with contextual commentary. Specific experimental numbers and methodological details of the paper should be verified against the linked PDF before citing quantitative conclusions.
Glossary
| Term | Meaning | |------|---------| | IR | Information Retrieval | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain, a ranking-quality metric | | RAG | Retrieval-Augmented Generation | | Agentic Search | Modeling search as sequential decision-making with tool calls | | Gen-IR | Generative Information Retrieval |