# LV-ROVER-MLT: Low-Resource Maltese OCR by Synthetic Fine-Tuning and Multi-Stream Arbitration

arXiv: https://arxiv.org/abs/2607.00250 · HTML: https://mlsanity.com/paper/2607.00250/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 26 references that are in the map.
- First posted on arXiv 30 Jun 2026.

## Builds on (most cited first, 8 of 8)

- [Synthetic Data for Text Localisation in Natural Images](https://mlsanity.com/paper/1604.06646/) (2016)
- [What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis](https://mlsanity.com/paper/1904.01906/) (2019)
- [SynthTIGER: Synthetic Text Image GEneratoR Towards Better Text Recognition Models](https://mlsanity.com/paper/2107.09313/) (2021)
- [Calamari - A High-Performance Tensorflow-based Deep Learning Package for Optical Character Recognition](https://mlsanity.com/paper/1807.02004/) (2018)
- [Pre-training Data Quality and Quantity for a Low-Resource Language: New Corpus and BERT Models for Maltese](https://mlsanity.com/paper/2205.10517/) (2022)
- [Morphological Analysis for the Maltese Language: The Challenges of a Hybrid System](https://mlsanity.com/paper/1703.08701/) (2017)
- [LV-ROVER: Lexicon Verified Recognizer Output Voting Error Reduction](https://mlsanity.com/paper/1707.07432/) (2017)
- [Synthetic Document Question Answering in Hungarian](https://mlsanity.com/paper/2505.23008/) (2025)

## Abstract (arXiv)

Maltese has substantial text corpora and pretrained language models, but paragraph-scale OCR training data remains scarce; NOMOCRAT provides 57 verified annotated pages. LV-ROVER-MLT combines synthetic fine-tuning of Tesseract~5 with five complementary recognition streams and lexicon-gated word-level arbitration adapted to Maltese diacritics and hyphenation. In the DocEng~2026 Maltese OCR competition, the system placed first with held-out CER 0.0074; the next-ranked submission scored 0.0161 and NOMOCRAT scored 0.0163. The same approach produced a significant improvement over stock Tesseract on Luxembourgish, while the Hungarian result was inconclusive. A 36,803-pair Maltese OCR corpus constructed from EUR-Lex and Wikipedia provides an additional paragraph-level resource. Code, model weights, and corpus data are public.