Text Adaptation to Plain Language and Easy Read via Automatic Post-Editing Cycles

2025-09-15

AI Summary

This study addresses the automatic simplification of Spanish texts into plain language and easy-read formats. We propose a two-stage approach integrating large language models (LLMs) with iterative post-editing. First, an LLM generates an initial simplified output; second, multiple rounds of automated post-editing refine the text, jointly optimizing readability (e.g., Flesch Reading Ease) and semantic fidelity (e.g., BERTScore) via dynamic threshold-based stopping criteria. This balances simplification efficacy and meaning preservation. Evaluated on the CLEARS international challenge, our method achieved first place in the plain-language task and second place in the easy-read task—demonstrating state-of-the-art performance in automated accessibility enhancement for Spanish textual content.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

We describe Vicomtech's participation in the CLEARS challenge on text adaptation to Plain Language and Easy Read in Spanish. Our approach features automatic post-editing of different types of initial Large Language Model adaptations, where successive adaptations are generated iteratively until readability and similarity metrics indicate that no further adaptation refinement can be successfully performed. Taking the average of all official metrics, our submissions achieved first and second place in Plain language and Easy Read adaptation, respectively.

Problem

• Automating text simplification for Plain Language
• Iterative post-editing cycles for readability improvement
• Optimizing readability and similarity metrics in adaptations

Innovation

• Automatic post-editing of LLM adaptations
• Iterative refinement cycles for readability
• Metrics-driven adaptation termination criteria