Enhancing table detection across many languages with masking
EMBLEM: Enhancing Multi-script Table Detection through Masking
Machine Learning
Summary
Table detection helps computers find tables in documents, but most methods work only well on English or Chinese text. The authors created a new dataset called MANDALA with tables in 18 languages and 15 scripts to study this problem better. They developed a method named EMBLEM that hides specific script and font details so the computer focuses on the overall layout, improving detection on multiple languages. This approach works better on the new dataset and still performs well on English documents. They share their data and code for others to use.
What this means in practice
- •For document analysis teams: Improve table detection systems to work accurately on documents with many languages and scripts without needing extensive multi-language training data.
- •For digital archiving specialists: Develop better tools for extracting tabular data from historical and multilingual documents across diverse fonts and scripts using script-agnostic layout features.
- •For enterprise software developers: Create commercial document processing software that reliably detects tables in multi-language environments by applying masking-based training methods.$Commercial implications: Enables selling advanced document analysis software to companies handling international documents by solving multi-script table detection challenges.
Authors
Dhruv Kudale, Udhay Brahmi, Ganesh Ramakrishnan
Abstract
Table detection is a core task in document analysis, supporting downstream applications such as information retrieval, document reconstruction, and visual question answering. While existing deep learning models perform well on English and Chinese documents, they struggle with multilingual, multi-script documents due to script diversity and the limited availability of labeled data. To address this challenge, we introduce MANDALA (Multi-script Annotated Documents for Table Detection), a manually curated dataset of 2,323 table-containing pages spanning 18 languages and 15 scripts across diverse domains. We also propose EMBLEM, a masking-based paradigm for Multi-script Table Detection (MTD). EMBLEM generates masked images that conceal script- and font-specific details, enabling models pre-trained on abundant English documents to focus on script-agnostic page layout. Experiments across three table detection architectures show that EMBLEM consistently outperforms strong baselines on MANDALA while remaining competitive on five standard English-dominant benchmarks. Using only English masked images for fine-tuning, with no multi-script training data, EMBLEM achieves an absolute F1-score gain of 20.8% on MANDALA. We release MANDALA along with the accompanying code and models at https://github.com/IITB-LEAP-OCR/EMBLEM.git.