Evaluation of Multilingual Text Simplification for the Mental Health Domain: Exploring Small Language Models

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing, Mental Health & Psychological Support · Depth: Advanced, quick

Summary

A study evaluated multilingual text simplification for the mental health domain, focusing on language-agnostic approaches using small language models (LMs). Researchers Olga Pelloni, Sandra Just, and Lars Bongo simplified ICD-11 articles on primary psychotic disorders across English, German, and French. The evaluation employed metrics for lexical complexity and readability, specifically the Measure of Textual Lexical Diversity (MTLD) and Flesch Reading Ease (FRE). Results indicated that acceptable simplified texts were produced exclusively in English. The study found that a joint analysis of MTLD and FRE offered the most comprehensive insight, effectively identifying optimal outcomes and different types of simplification issues. This preliminary research suggests challenges in extending text simplification beyond English for medical content.

Key takeaway

For NLP Engineers developing multilingual text simplification tools for healthcare, recognize that current small language models primarily yield acceptable results in English. You should prioritize English content for immediate deployment while investing in research to address the significant challenges observed in German and French. When evaluating simplification outputs, integrate both Measure of Textual Lexical Diversity (MTLD) and Flesch Reading Ease (FRE) for a comprehensive assessment of text quality and potential issues.

Key insights

Multilingual medical text simplification with small LMs is challenging, succeeding only in English for ICD-11 content.

Principles

• Text simplification is predominantly English-centric.
• Joint analysis of MTLD and FRE is effective for evaluation.
• Small LMs face limitations in multilingual medical simplification.

Method

The study simplified ICD-11 articles on primary psychotic disorders in English, German, and French using small LMs, evaluating lexical complexity and readability with MTLD and Flesch Reading Ease.

In practice

• Prioritize English for current LM-based medical simplification.
• Combine MTLD and FRE for robust text evaluation.
• Investigate language-specific challenges in LMs.

Topics

• Text Simplification
• Multilingual NLP
• Mental Health Domain
• Small Language Models
• Readability Metrics
• ICD-11

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Cross-Lingual Empirical Evaluation of Large Language Models for Arabic Medical Tasks — English-centric LLMs exhibit significant performance gaps and reliability issues in Arabic medical tasks.
• Annotation Guidelines and Challenges for Automatic Simplification of Portuguese Drug Leaflets — Simplifying complex drug leaflets via NLP improves patient comprehension and treatment adherence.
• Authorship Attribution in Multilingual Machine-Generated Texts — Multilingual authorship attribution for LLM-generated text is complex, with current monolingual methods showing limited cross-lingual transferability across diverse language families.
• Towards a Universal Dependencies Corpus for Portuguese Epidemiological Reports — Document extraction and representation significantly impact parsing epidemiological texts more than parser selection.
• SemEval-2026 Task 7: Everyday Knowledge Across Diverse Languages and Cultures — Evaluating LLM adaptability across diverse, low-resource language-culture pairs reveals critical performance and misalignment issues.
• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.

Open in AIssential →

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