A Larger Annotated Corpus of Portuguese Coreference

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

Summary

Researchers have introduced LLM-PREF, a new large language model (LLM) based system designed for coreference annotation in Portuguese texts. Coreference resolution is a vital natural language processing (NLP) task focused on identifying expressions that refer to the same entity within a text. The development of LLM-PREF addresses a significant challenge in Portuguese NLP: the scarcity of annotated coreference data. While LLM-PREF demonstrates rich world knowledge and strong inference capabilities, enabling it to recognize complex coreference patterns like pronominal anaphora, its performance did not surpass that of a previously established rule-based system during evaluation.

Key takeaway

For research scientists developing NLP systems for low-resource languages like Portuguese, you should carefully evaluate LLM-based annotation approaches against simpler, rule-based systems. While LLMs offer advanced inference, they may not always yield superior performance for specific tasks like coreference resolution, suggesting that hybrid or refined rule-based methods might still be more effective or efficient.

Key insights

LLMs can annotate Portuguese coreference, but a new LLM-based system did not outperform existing rule-based methods.

Principles

• Annotated data scarcity hinders NLP development.
• LLMs possess rich world knowledge and inference capacity.

Method

LLM-PREF, an LLM-based system, was proposed for Portuguese coreference annotation and evaluated against a prior rule-based system.

In practice

• Consider LLMs for complex linguistic patterns.
• Evaluate LLM performance against established baselines.

Topics

• Coreference Resolution
• Portuguese NLP
• Large Language Models
• LLM-PREF
• Annotated Corpus

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Textual Inference in Portuguese: Comparing Language Models — LLMs struggle with logic-sensitive semantic reasoning in Portuguese, highlighting the need for specialized benchmarks.
• LLMs for automatic annotation of Mandarin narrative transcripts — LLMs can automate complex discourse-level annotation in non-English speech, significantly reducing time.
• Coreference annotation for the characterization of characters in Portuguese literary works — Coreference annotation enhances character characterization in Portuguese literary corpora for computational literary studies.
• Linguistic Issues in Computational Linguistics for Non-English Languages — English-centric LLM design creates systematic linguistic weaknesses for non-English languages across multiple levels.
• A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research — A RoBERTa classifier, trained on LLM-generated annotations, accurately distinguishes manner and result verbs at scale.
• TalkTag: Fine-Grained Morphosyntactic Error Annotation for Transcribed Speech — TalkTag automates fine-grained morphosyntactic error annotation in spoken transcripts using an LLM, even with scarce data.

Open in AIssential →

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