• Home
  • Search
  • Combination of multiple aligned recognition outputs using WFST and LSTM
  • Open Access IconOpen Access
  • Cite Icon10
  • https://doi.org/10.1109/icdar.2015.7333720Copy DOI Icon

Combination of multiple aligned recognition outputs using WFST and LSTM

  • Aug 1, 2015
  • Mayce Al Azawi +2 more
Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

The contribution of this paper is a new strategy of integrating multiple recognition outputs of diverse recognizers. Such an integration can give higher performance and more accurate outputs than a single recognition system. The problem of aligning various Optical Character Recognition (OCR) results lies in the difficulties to find the correspondence on character, word, line, and page level. These difficulties arise from segmentation and recognition errors which are produced by the OCRs. Therefore, alignment techniques are required for synchronizing the outputs in order to compare them. Most existing approaches fail when the same error occurs in the multiple OCRs. If the corrections do not appear in one of the OCR approaches are unable to improve the results. We design a Line-to-Page alignment with edit rules using Weighted Finite-State Transducers (WFST). These edit rules are based on edit operations: insertion, deletion, and substitution. Therefore, an approach is designed using Recurrent Neural Networks with Long Short-Term Memory (LSTM) to predict these types of errors. A Character-Epsilon alignment is designed to normalize the size of the strings for the LSTM alignment. The LSTM returns best voting, especially when the heuristic approaches are unable to vote among various OCR engines. LSTM predicts the correct characters, even if the OCR could not produce the characters in the outputs. The approaches are evaluated on OCR's output from the UWIII and historical German Fraktur dataset which are obtained from state-of-the-art OCR systems. The experiments shows that the error rate of the LSTM approach has the best performance with around 0.40%, while other approaches are between 1.26% and 2.31%.

Similar Papers
  • Book Chapter
  • Citations8

Character-Level Alignment Using WFST and LSTM for Post-processing in Multi-script Recognition Systems - A Comparative Study

  • Jan 01, 2014
  • Mayce Al Azawi +3
  • Research Article
  • Citations5

Robust named entity detection from optical character recognition output

  • Apr 13, 2011
  • International Journal on Document Analysis and Recognition (IJDAR)
  • Krishna Subramanian +2
  • Conference Article
  • Citations54

A hierarchical, HMM-based automatic evaluation of OCR accuracy for a digital library of books

  • Jun 11, 2006
  • Shaolei Feng +1
  • Research Article
  • Citations39

Unsupervised profiling of OCRed historical documents

  • Nov 09, 2012
  • Pattern Recognition
  • Ulrich Reffle +1
  • Conference Article
  • Citations7

Improving Optical Character Recognition(OCR) Accuracy using Multi-Layer Perceptron(MLP)

  • Apr 11, 2023
  • M Krishnamoorthi +3
  • Conference Article
  • Citations1

<title>Identification and correction of rejection and substitution errors in optical character recognition systems</title>

  • Apr 14, 1993
  • Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE
  • Glenn S Himes +2
  • Conference Article
  • Citations1

Benchmarking commercial OCR engines for technical drawings indexing

  • Sep 10, 2001
  • J.C Lecoq +3
  • Book Chapter
  • Citations8

Soft Computing Techniques for Optical Character Recognition Systems

  • Dec 24, 2016
  • Arindam Chaudhuri +3
  • Research Article
  • Citations7

Applicability of OCR Engines for Text Recognition in Vehicle Number Plates, Receipts and Handwriting

  • Dec 01, 2023
  • Journal of Circuits, Systems and Computers
  • Utsav Poudel +3
  • PDF
  • Research Article
  • Citations7

An OCR Engine for Printed Receipt Images using Deep Learning Techniques

  • Jan 01, 2023
  • International Journal of Advanced Computer Science and Applications
  • Cagri Sayallar +2
  • Conference Article
  • Citations8

Error Correction with In-domain Training across Multiple OCR System Outputs

  • Sep 01, 2011
  • William B Lund +1
  • Conference Article

Treating OCR Output as a Language (TOOL) – Improving OCR Output with Seq2Seq Translation

  • Oct 15, 2025
  • Thomas Asselborn +3
  • Research Article
  • Citations4

Book History in the Early Modern OCR Project, or, Bringing Balance to the Force

  • Jan 01, 2013
  • Journal for Early Modern Cultural Studies
  • Jacob Heil +1
  • Conference Article
  • Citations24

A Study on the Accuracy of OCR Engines for Source Code Transcription from Programming Screencasts

  • Jun 29, 2020
  • Abdulkarim Malkadi +2
  • Book Chapter
  • Citations5

JPEG for Arabic Handwritten Character Recognition: Add a Dimension of Application

  • Oct 01, 2008
  • Abdurazzag Ali +1
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.