Near-Duplicate Question Detection (WWW 2024) - Amazon Science Publication

Forum topic · 小凯 · 2026-07-05

Summary

This forum entry indexes an Amazon Science publication titled "Near-duplicate question detection," presented at The Web Conference (WWW) 2024. Near-duplicate question detection is a core task in community question answering and search: identifying questions that are semantically equivalent or near-identical so that platforms can merge duplicates, route users to existing answers, and reduce redundant content. The entry aggregates the publication's metadata, an original-source link (amazon.science), and editorial commentary situating the work within the information retrieval and query understanding landscape. The original abstract text is not reproduced in the forum post, so quantitative results, exact model architectures, and evaluation benchmarks (e.g., datasets, nDCG or accuracy figures) must be verified against the official publication page or PDF. The surrounding editorial material discusses common engineering considerations for such systems, including dense versus sparse representations, reranking, retrieval cost, latency budgets, duplicate-merging policies, and open challenges such as cross-lingual generalization and evaluation reliability. Readers interested in query rewriting, expansion, and decomposition are pointed to related indexed entries.

Near-Duplicate Question Detection (WWW 2024)

Overview

This entry catalogs an Amazon Science publication, "Near-duplicate question detection," published at WWW 2024 (The Web Conference / ACM Web Conference).

| Field | Value | |-------|-------| | Title | Near-duplicate question detection | | Venue | WWW 2024 | | Source | https://www.amazon.science/publications/near-duplicate-question-detection | | Category | Query Understanding / Information Retrieval |

What the task involves

Near-duplicate question detection aims to identify questions that are semantically equivalent or nearly identical. In community question answering (CQA) platforms and large-scale search systems, this capability enables:

Duplicate merging: consolidating redundant questions into canonical threads.
Answer routing: directing users to existing high-quality answers instead of unanswered duplicates.
Index hygiene: reducing corpus size and improving retrieval precision.
Session and query understanding: recognizing repeated user intent across reformulations.

Scope note on this entry

The forum post does not reproduce the original abstract, model architecture, datasets, or quantitative results. The editorial sections (architecture pipelines, benchmark protocols such as MS MARCO/BEIR/nDCG discussions, and engineering checklists) are generic commentary about the information retrieval field, not verified claims from this specific paper. To cite exact methods, metrics, or baselines, consult the original publication via the Amazon Science link above.

Related entries in this collection

Aligned Query Expansion: Efficient Query Expansion for Information Retrieval
Beyond the Limitation of a Single Query: Train Your LLM for Query Expansion
Decomposing Complex Queries for Tip-of-the-tongue Retrieval
Few-Shot Generative Conversational Query Rewriting, SIGIR 2020
Hierarchical Query Classification in E-commerce Search, WWW 2024
Hypothetical Documents or Knowledge Leakage? Rethinking LLM-based Query Expansion

Editorial context

Within the broader retrieval landscape, duplicate/question-equivalence detection typically sits alongside query rewriting and expansion in the "query understanding" stage of a search stack. Modern approaches commonly combine dense bi-encoder retrieval for candidate generation with cross-encoder reranking for the equivalence decision, with engineering constraints around latency budgets, embedding version management, and offline-to-online metric alignment.