Data Augmentation for Named Entity Recognition in Domain-Specific Scenarios in Portuguese

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

Summary

This study evaluates data augmentation techniques for Named Entity Recognition (NER) in domain-specific Portuguese datasets, addressing the challenge of limited labeled data in low-resource languages. Researchers applied rule-based methods, back-translation, and large language models (LLMs) to augment four distinct datasets. These augmented datasets were then used to train Transformer-based NER models. The findings indicate that most augmentation techniques enhanced model performance compared to the baseline, with the best improvements observed when using PP-LLM, SR, and MR methods.

Key takeaway

For research scientists developing NER models in low-resource languages like Portuguese, integrating data augmentation techniques is crucial. You should prioritize methods such as PP-LLM, SR, and MR to enhance model performance and reduce reliance on extensive human-annotated data, thereby accelerating development cycles.

Key insights

Data augmentation significantly improves NER performance in low-resource, domain-specific Portuguese datasets.

Principles

• Data augmentation mitigates annotation costs.
• Transformer models benefit from augmented data.

Method

The study employed rule-based, back-translation, and LLM-based data augmentation on four Portuguese NER datasets, then trained Transformer models to evaluate performance improvements.

In practice

• Use PP-LLM for NER data augmentation.
• Consider SR and MR techniques for better results.

Topics

• Named Entity Recognition
• Data Augmentation
• Portuguese Language Processing
• Domain-Specific NER
• Transformer Models

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

Related on AIssential

• Data Augmentation for Historical NER: A Systematic Comparison of Lexical and LLM-based Approaches — This paper systematically compares lexical and LLM-based data augmentation for Historical Named Entity Recognition.
• Negation-Aware Data Augmentation for Portuguese Natural Language Inference — Targeted negation-aware data augmentation improves BERT-based NLI model performance on Portuguese datasets by reducing negation bias.
• Low-resource Language Discrimination Towards Chinese Dialects with Transfer learning and Data Augmentation — Transfer learning and data augmentation effectively overcome scarce annotation resources for low-resource Chinese dialect discrimination.
• spaCy’s entity recognition model: incremental parsing with Bloom embeddings & residual CNNs — spaCy v2's NER model leverages transition-based parsing, Bloom embeddings, and residual CNNs, underscoring the need for domain-specific training data.
• How to Leverage NER and Advanced NLP Techniques for Life Sciences — NER and advanced NLP are essential for structuring and interpreting the vast unstructured text data in Life Sciences.
• Retrieval-Augmented Self-Taught Reasoning Model with Adaptive Chain-of-Thought for ASR Named Entity Correction — RASTAR improves ASR named entity correction by combining robust NER with adaptive, self-taught reasoning.

Open in AIssential →

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