Building an efficient OCR system for historical documents with little training data

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F20%3A43958971" target="_blank" >RIV/49777513:23520/20:43958971 - isvavai.cz</a>

Výsledek na webu

<a href="https://link.springer.com/content/pdf/10.1007/s00521-020-04910-x.pdf" target="_blank" >https://link.springer.com/content/pdf/10.1007/s00521-020-04910-x.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/s00521-020-04910-x" target="_blank" >10.1007/s00521-020-04910-x</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Building an efficient OCR system for historical documents with little training data

Popis výsledku v původním jazyce

As the number of digitized historical documents has increased rapidly it is necessary to provide efficient methods of information retrieval and knowledge extraction to make the data accessible. Such methods are dependent on optical character recognition (OCR) which converts the document images into textual representations. This paper introduces a set of methods that allows performing an OCR on historical document images using only a small amount of real, manually annotated training data. The presented OCR system includes two main tasks: page layout analysis including text block and line segmentation and OCR. Our seg-mentation methods are based on fully convolutional networks, and the OCR approach utilizes recurrent neural networks. We show that both the segmentation and OCR tasks are feasible with only a few annotated real data samples. The experiments aim at determining the best way how to achieve good performance with the given small set of data. We also demonstrate that obtained scores are comparable or even better than the scores of several state-of-the-art systems.

Název v anglickém jazyce

Building an efficient OCR system for historical documents with little training data

Popis výsledku anglicky

As the number of digitized historical documents has increased rapidly it is necessary to provide efficient methods of information retrieval and knowledge extraction to make the data accessible. Such methods are dependent on optical character recognition (OCR) which converts the document images into textual representations. This paper introduces a set of methods that allows performing an OCR on historical document images using only a small amount of real, manually annotated training data. The presented OCR system includes two main tasks: page layout analysis including text block and line segmentation and OCR. Our seg-mentation methods are based on fully convolutional networks, and the OCR approach utilizes recurrent neural networks. We show that both the segmentation and OCR tasks are feasible with only a few annotated real data samples. The experiments aim at determining the best way how to achieve good performance with the given small set of data. We also demonstrate that obtained scores are comparable or even better than the scores of several state-of-the-art systems.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

O - Projekt operacniho programu

Rok uplatnění

2020

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Neural Computing and Applications

ISSN

0941-0643

e-ISSN

—

Svazek periodika

32

Číslo periodika v rámci svazku

23

Stát vydavatele periodika

GB - Spojené království Velké Británie a Severního Irska

Počet stran výsledku

19

Strana od-do

17209-17227

Kód UT WoS článku

000531222300001

EID výsledku v databázi Scopus

2-s2.0-85084519412