Constructing a Silver Corpus for Weakly Supervised Vietnamese Event Extraction using Cross-Document N-ary Relation Filtering

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Natural Language Processing · Depth: Expert, quick

Summary

A weakly supervised framework addresses the scarcity of annotated data for Vietnamese event extraction, a significant challenge for low-resource languages. This approach constructs a silver corpus through pseudo-labeling, incorporating a cross-document n-ary relation filtering strategy to minimize noise by leveraging consistency across multiple articles describing the same event. It further enhances data diversity using schema-based augmentation. Experiments on the BKEE benchmark demonstrated consistent improvements, validating the framework's effectiveness. The data is publicly available at https://github.com/Larken1612/VietEE2.

Key takeaway

For NLP Engineers building event extraction systems in low-resource languages, this framework offers a practical solution to data scarcity. You should consider implementing cross-document n-ary relation filtering to improve pseudo-label quality and integrate schema-based augmentation to diversify your training data, as demonstrated by its effectiveness on the BKEE benchmark. This can significantly enhance system performance.

Key insights

Cross-document consistency and schema-based augmentation improve weakly supervised event extraction for low-resource languages.

Principles

• Cross-document consistency reduces pseudo-label noise.
• Schema-based augmentation enhances data diversity.

Method

Construct a silver corpus via pseudo-labeling, apply cross-document n-ary relation filtering for noise reduction, and use schema-based augmentation for diversity.

In practice

• Implement cross-document checks for noisy labels.
• Apply schema-based augmentation for data diversity.

Topics

• Vietnamese Event Extraction
• Weakly Supervised Learning
• Silver Corpus
• Pseudo-labeling
• Cross-Document Filtering
• Data Augmentation
• Low-Resource Languages

Code references

• Larken1612/VietEE2

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.
• Closing the Last Mile in Document AI: Improving Extraction Quality in Azure Content Understanding — Labeled examples significantly boost document field extraction accuracy in generative AI systems like Azure Content Understanding.
• A BART-based approach with hierarchical strategy for Vietnamese abstractive multi-document summarization — A hierarchical, BART-based approach for Vietnamese multi-document summarization uses "golden summaries" to guide document shortening, preventing information loss.
• Grammar of the Wave: Towards Explainable Multivariate Time Series Event Detection via Neuro-Symbolic VLM Agents — Neuro-symbolic VLM agents utilize Event Logic Trees for explainable, zero-shot time series event detection from natural language descriptions.
• CIAN: Multi-Stage Framework for Event-Enriched Image Captioning via Retrieval-Augmented Generation — CIAN enriches image captions with external event context using retrieval-augmented generation and linguistic refinement.
• Few-Shot Biomedical Relation Extraction with Large Language Models: A Viable Alternative to Supervised Learning? — LLMs offer a viable few-shot alternative for biomedical relation extraction, outperforming prior few-shot methods, especially for rare relations.

Open in AIssential →

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