Improving Conversational Passage Re-ranking with View Ensemble (SIGIR 2023 Short Paper)
Overview
This forum entry indexes a SIGIR 2023 short paper:
- Title: Improving Conversational Passage Re-ranking with View Ensemble
- Venue: SIGIR 2023 (short paper)
- Source: https://dl.acm.org/doi/abs/10.1145/3539618.3592002
- Topic area: Conversational Search
- Understanding user intent across multiple turns (context-dependent queries)
- Balancing latency budgets against re-ranking depth
- Ensuring evaluation reliability when offline metrics diverge from online user satisfaction
- Handling hallucination and safety risks when generation is coupled with retrieval
- A Survey of Conversational Search (Sep 2025, ACM)
- Engineering Conversational Search Systems: A Review of Applications
- Agentic Conversational Search with Contextualized Reasoning via Reinforcement Learning
- CTR-Guided Generative Query Suggestion in Conversational Search (EMNLP)
- ChatRetriever: Adapting Large Language Models for Generalized and Robust Retrieval
- CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models
- Original paper: *Improving Conversational Passage Re-ranking with View Ensemble*, SIGIR 2023 short paper. DOI: 10.1145/3539618.3592002
The paper targets conversational passage re-ranking, a core component of conversational search systems where documents must be re-ordered according to the information need expressed over multiple dialogue turns.
> Note: The original forum post does not reproduce the paper's full abstract or experimental tables. Quantitative results, baselines, and ablations should be verified against the original PDF at the ACM link above.
Context in Conversational Search
Modern search stacks typically follow a cascade of retrieval → re-ranking → generation. In conversational settings, additional challenges arise:
The view ensemble approach referenced in the title suggests combining multiple representations or "views" of the query/document context to improve re-ranking decisions.
Related Work Cross-references
The post links this paper with adjacent entries:
Key Takeaways for Practitioners
1. Architecture: Cascade retrieval + re-ranking + generation remains mainstream; agentic paradigms increasingly treat retrieval strategy itself as learnable. 2. Evaluation: Offline metrics (nDCG@10, MRR) increasingly diverge from online satisfaction; LLM-as-judge should be cross-validated with human assessment. 3. Deployment constraints: Latency, cost, interpretability, and safety are hard constraints in production, beyond academic benchmarks. 4. Reproduction advice: Verify reported numbers against the original paper; check for statistical significance and computational cost reporting.
Glossary
| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | nDCG | Normalized Discounted Cumulative Gain, a ranking quality metric | | Agentic Search | Modeling search as sequential decision-making and tool invocation |