Comparing Machine Learning Approaches for Table Recognition in Historical Register Books

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 modelled, 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.

Matching Table Structures of Historical Register Books using Association Graphs

“…The table rows are detected and evaluated in a second step (see Section III-B). The dataset will be published as part of a competition dataset on table recognition planned in 2018 within the context of the READ project 5 . The GT is represented as extended PAGE XML (see https://github.com/ Transkribus/TranskribusPageformat).…”

“…In this section, the table structure matching using association graphs is presented. As a second part, the work of Clinchant et al [5] is shortly summarized which allows for row detection based on detected columns of the presented table matching and detected baselines.…”

“…Based on the columns and the header of the matched table, the table rows are detected using baseline information of the basic layout analysis [14]. The detection of the rows is presented in [5] and summarized in the following section.…”

“…After the detection of the table structure, a baseline detection can be performed and the baselines are attributed to the according cells (also split if necessary). The row detection is based on Clinchant et al [5] which detects rows using the baseline information. The presented methodology is evaluated on a historical dataset, which will be published as part of a competition dataset on table recognition planned in 2018 within the context of the READ project 2 .…”

Transforming scholarship in the archives through handwritten text recognition

“…Passau Diocesan Archives are also working to improve the automated Layout Analysis of tabular data by sorting tables from their collection into different categories that can be used as training templates. Improved table recognition and the possibility of exporting tabular data will have significant implications for the field, since many archival documents are laid out in tables and forms (Clinchant et al, 2018). The National Archives of Finland [27] would like to enhance the usability of their vast collections of governmental records, many of which are digitised but not transcribed.…”