RALS: Resources and Baselines for Romanian Automatic Lexical Simplification

2026-07-22 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

The RALS project introduces the first comprehensive dataset for Romanian Automatic Lexical Simplification (LS) and Lexical Complexity Prediction (LCP). This initiative provides human lexical complexity annotations for 3,921 word samples in context. It also proposes a novel methodology for ordering simplification suggestions, utilizing a pairwise ranking approximation method based on separate human judgments to arrange candidates from simple to complex. Furthermore, RALS explores several new pipelines for complexity prediction and simplification, culminating in the presentation of the first dedicated text simplification system for the Romanian language. This work establishes crucial baselines and resources for future research in Romanian NLP.

Key takeaway

For NLP engineers and researchers focusing on low-resource languages, particularly Romanian, RALS offers foundational resources. If you are developing text simplification systems for Romanian, you should consider integrating the RALS dataset and its proposed methodologies. This work provides essential baselines and a robust framework, accelerating your ability to build and evaluate effective lexical simplification solutions for the language.

Key insights

RALS provides the first integrated dataset and system for Romanian lexical simplification and complexity prediction.

Principles

• Pairwise ranking improves simplification suggestion order.
• Human judgments are key for ranking simplification candidates.

Method

A methodology orders simplification suggestions using pairwise ranking approximation, arranging candidates from simple to complex based on human judgments.

In practice

• Develop lexical simplification systems for Romanian.
• Apply pairwise ranking to candidate ordering tasks.

Topics

• Romanian NLP
• Lexical Simplification
• Lexical Complexity Prediction
• Text Simplification
• Dataset
• Low-Resource Languages

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• RALS: Resources and Baselines for Romanian Automatic Lexical Simplification — RALS provides the first Romanian lexical simplification resources, with a hybrid rule-based system outperforming LLMs for this low-resource language.
• LegalSim-PT: Building a Dataset for Legal Document Simplification in Portuguese Leveraging Linguistic Metrics — LegalSim-PT is the first large-scale Portuguese legal document simplification dataset, created using linguistic metrics and data augmentation.
• Cross-lingual Relation Extraction with Large Language Models: Zero-Shot, Few-Shot, and Fine-Tuned Evaluation on Romanian — QLoRA fine-tuning significantly boosts cross-lingual relation extraction performance for low-resource languages, outperforming prompt-only LLMs.
• Abstractiveness Metrics for Evaluating Text Summarization: A Refined Formulation with Empirical Validation — New metrics (RA, SA, AR) quantify summary abstractiveness, distinguishing extractive from abstractive models and flagging potential hallucination.
• Failure at BEA 2026 Shared Task 1: One Pipeline, Three L1s: A Unified Language-Agnostic System for Vocabulary Difficulty Prediction — A unified, language-agnostic system can effectively predict vocabulary difficulty across diverse L1s without language-specific adaptation.
• RoLegalGEC: Legal Domain Grammatical Error Detection and Correction Dataset for Romanian — RoLegalGEC provides the first Romanian legal grammatical error dataset and evaluates neural models for its use.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.