Comparing Machine Learning Approaches for Table Recognition in Historical Register Books

TL;DR

This paper compares two machine learning methods, Conditional Random Fields and Graph Convolutional Networks, for recognizing tables in handwritten historical registry books, achieving similar high accuracy suitable for information extraction.

Contribution

It introduces a comparative analysis of ML approaches for table recognition in handwritten historical documents, with open source tools and datasets.

Findings

Both methods achieved an 89 F1 score.

Methods are suitable for information extraction.

Open source software and datasets provided.

Abstract

We present in this paper experiments on Table Recognition in hand-written registry books. We first explain how the problem of row and column detection is modeled, and then compare two Machine Learning approaches (Conditional Random Field and Graph Convolutional Network) for detecting these table elements. Evaluation was conducted on death records provided by the Archive of the Diocese of Passau. Both methods show similar results, a 89 F1 score, a quality which allows for Information Extraction. Software and dataset are open source/data.