Moby Dick meets GEOCR: lexical considerations in word recognition

Published 22 November 2002

Andreas Spitz

DOI10.1109/icdar.1997.619845

Citations9

TL;DR

The effects of lexical content, structure and processing on the performance of GEOCR as a word recognition engine are described, describing the recognition of a particular text, Moby Dick.

Abstract

The author has previously (Proc. Int. Conf. on Doc. Anal. and Recognition, Montreal, pp. 723-728, 1995) described a high-speed, lexically driven OCR called GEOCR (Good Enough Optical Character Recognition). This paper expands on that work by describing the effects of lexical content, structure and processing on the performance of GEOCR as a word recognition engine, describing the recognition of a particular text, Moby Dick. Word recognition performance is shown to be enhanced by the application of an appropriate lexicon. Recognition speed is essentially independent of the details of lexical content, provided that the intersection of the occurrences of words in the document and the lexicon is high. Word recognition accuracy is dependent on both the intersection and specificity of the lexicon.

Keywords

Computer Science

The Fourth Annual Test of OCR Accuracy

144 Citations1995Stephen V. Rice, Frank R. Jenkins +1 more

The annual test of optical character recognition systems known as “page readers” accepts as input a bitmapped image of any document page, and attempts to identify the machine-printed characters on the page.

Proceedings of the IEEEDocument analysis-from pixels to contents

64 Citations1992Jürgen Schürmann, Norbert Bartneck +4 more

The authors present a conceptual framework for solving the task of document analysis, which, in essence, consists in the conversion of the document's pixel representation into an equivalent knowledge network representation holding the document"s content and layout.

IEEE Transactions on Pattern Analysis and Machine IntelligenceHybrid contextural text recognition with string matching

51 Citations1993R.M.K. Sinha, B. Prasada +2 more

The hybrid contextural algorithm for reading real-life documents printed in varying fonts of any size is presented and word-level hypotheses are generated using hybrid contextual text processing.

IEEE Transactions on Pattern Analysis and Machine IntelligenceValidation of image defect models for optical character recognition

45 Citations1996Yanhong Li, Daniel Lopresti +2 more

A defect model is validated if the OCR errors induced by the model are indistinguishable from the errors encountered when using real scanned documents, as well as four measures to quantify this similarity.

IEEE Transactions on Pattern Analysis and Machine IntelligenceDecoding Substitution Ciphers by Means of Word Matching with Application to OCR

44 Citations1987George Nagy, Sharad Seth +1 more

The heuristic backtrack algorithm typically visits only a few hundred among the 26! possible nodes on sample texts ranging from 100 to 600 words, thereby opening up the possibility of automatic translation of a scanned document into a standard character code, such as ASCII.

Pattern RecognitionOn partitioning a dictionary for visual text recognition

25 Citations1990R.M.K. Sinha

This paper has used word-length, word-envelope, character combination and a combination of these as the basis of partitioning the dictionary, and considered approximately 10,000 of the most frequently used words for analysis and comparison.

An OCR based on character shape codes and lexical information

23 Citations2002Andreas Spitz

An OCR process which has as its principal attributes high speed of operation and tunability to the lexical content of the documents to which it is applied, and relies on the structure of English and the high percentage of singleton mappings between the shape codes and the characters in the words.

Transmedia machine and its keyword search over image texts

19 Citations1988Yasuaki Tanaka, Hajime Torii

Systeme de gestion personnel d'information automatisee pouvant traiter aussi bien le texte que l'image, encore au stade de prototype est decrite a l'aide d'exemples.

Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE<title>Text characterization by connected component transformations</title>

16 Citations1994Larry Spitz

A method of handling text lines from several different languages that is robust in the presence of common printing and scanning artifacts is developed and incorporates a computationally inexpensive method for straightening curved lines and segmenting kerned characters.