Structured Summaries for Retrieval-Augmented Generation in Portuguese-Language Consumer Complaints

2026-04-12 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, medium

Summary

A study presented at PROPOR 2026 by Sant'Ana et al. investigates the impact of schema-guided structured summaries on dense retrieval within Retrieval-Augmented Generation (RAG) systems, specifically for Portuguese-language consumer complaints. The research compares embeddings derived from raw, lengthy, and noisy interaction texts against those from LLM-generated structured summaries. In a controlled evaluation, summary-based retrieval achieved a Recall@1 of 0.527, a significant improvement over the 0.001 observed when indexing raw interactions. Furthermore, summary-based retrieval reached a Recall@10 of 0.610, demonstrating gains exceeding two orders of magnitude. These findings indicate that structured summaries substantially enhance retrieval effectiveness and reliability, particularly at low cutoffs, making them highly suitable for RAG pipelines in challenging text environments.

Key takeaway

For AI Architects designing RAG systems for domains with verbose or unstructured text, such as customer support logs or legal documents, you should prioritize integrating LLM-generated structured summaries into your retrieval pipeline. This approach can dramatically improve retrieval accuracy at low cutoffs, as demonstrated by a Recall@1 of 0.527 versus 0.001 for raw text, leading to more reliable and relevant augmented generation outputs.

Key insights

Structured summaries significantly improve dense retrieval effectiveness in RAG systems for noisy, lengthy texts.

Principles

• Document representation is critical for dense retrieval.
• Noisy, lengthy texts limit retrieval effectiveness.

Method

The method involves comparing embeddings from raw texts versus LLM-generated structured summaries in a controlled evaluation on Portuguese consumer complaints to assess dense retrieval performance.

In practice

• Use LLM-generated summaries for RAG indexing.
• Apply schema-guided summaries to noisy text data.

Topics

• Retrieval-Augmented Generation
• Dense Retrieval
• Structured Summaries
• LLM-generated Summaries
• Consumer Complaints

Best for: AI Architect, AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Baseline Enterprise RAG, From PDF to Highlighted Answer — A minimal RAG pipeline can provide verifiable, sourced answers by structuring outputs and linking directly to document evidence.
• Evaluation of Chunking Strategies for Effective Text Embedding in Low-Resource Language on Agricultural Documents — Effective text chunking significantly improves RAG performance in low-resource, morphologically complex languages.
• RARE: Redundancy-Aware Retrieval Evaluation Framework for High-Similarity Corpora — Existing RAG benchmarks misrepresent real-world performance due to high document redundancy and similarity in enterprise corpora.
• Disco-RAG: Discourse-Aware Retrieval-Augmented Generation — Explicitly modeling discourse structure and planning significantly improves RAG system performance, coherence, and factual consistency.
• Retrieval-Augmented Generation as a Bridge Between Classical NLP Pipelines and Modern LLM Systems — Retrieval-augmented generation (RAG) addresses the limitation of static knowledge in large language models (LLMs) by connecting them to external, up-to-date knowledge sources.
• Retrieval Augmented Generation (RAG) system — RAG systems enhance LLMs by integrating external knowledge retrieval, mitigating hallucination and outdated information.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.