Document-Level Text Simplification in Estonian Using Large Language Models
Despite advances in sentence-level simplification for high-resource languages, document-level simplification in morphologically rich, low-resource languages such as Estonian remains largely unexplored.
ProofPaper ↗
Key points
- Document-level text simplification involves transformations that go beyond sentence-internal edits, addressing discourse coherence, anaphora resolution, and cross-paragraph consistency.
- This study presents a comprehensive evaluation of five state-of-the-art multilingual large language models (LLMs) for document-level simplification in Estonian.
- The evaluation framework integrates automatic metrics assessing readability, semantic preservation, and discourse coherence, alongside a structured manual annotation protocol.
- This work contributes novel document-level coherence metrics, evidence-based prompting strategies, and publicly available resources for reproducibility.
Sources (1)
- [1]Document-Level Text Simplification in Estonian Using Large Language ModelsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 04:38 PM
Despite advances in sentence-level simplification for high-resource languages, document-level simplification in morphologically rich, low-resource languages such as Estonian remains largely unexplored.
Document-level text simplification involves transformations that go beyond sentence-internal edits, addressing discourse coherence, anaphora resolution, and cross-paragraph consistency.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 7, 2026Google rolls out improved SynthID AI content detector, now available globally
- Oct 6, 2026EmbeddingGemma 2: an open, lightweight multimodal embedding model
- Oct 3, 2026[AINews] not much happened today
- Oct 2, 2026[AINews] Pi 1.0, Pi Durable, and AIE NYC
- Sep 30, 2026googleapis/python-genai v2.26.0
- Sep 30, 2026Gemini 4 Argon: our next era of frontier intelligence