QAEA-DR: A Unified Text Augmentation Framework for Dense Retrieval

2024-07-29

IEEE Transactions on Knowledge and Data Engineering

AI Summary

To address information loss from long-text embeddings and inaccurate matching of low-quality documents in dense retrieval, this paper proposes a unified text enhancement framework that requires no modification to encoders or retrievers. The method introduces a novel enhancement paradigm integrating zero-shot question generation with element-driven event extraction, coupled with an LLM-based iterative re-generation mechanism guided by dynamic scoring—significantly improving the information density and relevance of enhanced texts. Entirely grounded in zero-shot prompting of large language models (LLMs), it avoids fine-tuning or additional annotations. Evaluated on multiple standard dense retrieval benchmarks, the approach achieves substantial improvements in recall (e.g., +2.1% @10 on MSMARCO Dev). Theoretical analysis and empirical validation confirm its effectiveness and seamless compatibility with existing retrieval systems, establishing a lightweight, general-purpose, and efficient pre-retrieval enhancement paradigm.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

In dense retrieval, embedding long texts into dense vectors can result in information loss, leading to inaccurate query-text matching. Additionally, low-quality texts with excessive noise or sparse key information are unlikely to align well with relevant queries. Recent studies mainly focus on improving the sentence embedding model or retrieval process. In this work, we introduce a novel text augmentation framework for dense retrieval. This framework transforms raw documents into information-dense text formats, which supplement the original texts to effectively address the aforementioned issues without modifying embedding or retrieval methodologies. Two text representations are generated via large language models (LLMs) zero-shot prompting: question-answer pairs and element-driven events. We term this approach QAEA-DR: unifying question-answer generation and event extraction in a text augmentation framework for dense retrieval. To further enhance the quality of generated texts, a scoring-based evaluation and regeneration mechanism is introduced in LLM prompting. Our QAEA-DR model has a positive impact on dense retrieval, supported by both theoretical analysis and empirical experiments.

Problem

• Addresses information loss in dense text embeddings.
• Improves alignment of low-quality texts with queries.
• Enhances dense retrieval without altering existing methodologies.

Innovation

• Text augmentation framework for dense retrieval
• Generates question-answer pairs and event-driven texts
• Scoring-based evaluation enhances text quality