Computer Vision and Image Processing: Fundamentals and Applications

183 Pages·2019·10.522 MB·English

by  Manas Kamal Bhuyan

Document Processing
Using Machine Learning

Document Processing
Using Machine Learning
Edited by
Sk Md Obaidullah, KC Santosh,
Teresa Gonçalves, Nibaran Das
and Kaushik Roy
CRC Press
Taylor & Francis Group
52 Vanderbilt Avenue,
New York, NY 10017
© 2020 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed on acid-free paper
International Standard Book Number-13: 978-0-367-21847-8 (Hardback)
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors
and publishers have attempted to trace the copyright holders of all material reproduced in this
publication and apologize to copyright holders if permission to publish in this form has not been
obtained. If any copyright material has not been acknowledged please write and let us know so we
may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known
or hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Contents
Preface ....................................................................................................................vii
Editors ......................................................................................................................ix
Contributors .........................................................................................................xiii
1. Artificial Intelligence for Document Image Analysis ............................1
Himadri Mukherjee, Payel Rakshit, Ankita Dhar, Sk Md Obaidullah,
KC Santosh, Santanu Phadikar and Kaushik Roy
2. An Approach toward Character Recognition of Bangla
Handwritten Isolated Characters ..............................................................15
Payel Rakshit, Chayan Halder and Kaushik Roy
3. Artistic Multi-Character Script Identification ........................................29
Mridul Ghosh, Himadri Mukherjee, Sk Md Obaidullah, KC Santosh,
Nibaran Das and Kaushik Roy
4. A Study on the Extreme Learning Machine and Its
Applications ...................................................................................................43
Himadri Mukherjee, Sahana Das, Subhashmita Ghosh,
Sk Md Obaidullah, KC Santosh, Nibaran Das and Kaushik Roy
5. A Graph-Based Text Classification Model for Web Text
Documents .....................................................................................................53
Ankita Dhar, Niladri Sekhar Dash and Kaushik Roy
6. A Study of Distance Metrics in Document Classification ...................69
Ankita Dhar, Niladri Sekhar Dash and Kaushik Roy
7. A Study of Proximity of Domains for Text Categorization .................85
Ankita Dhar, Niladri Sekhar Dash and Kaushik Roy
8. Supervised Learning for Aggression Identification and
Author Profiling over Twitter Dataset ....................................................101
Kashyap Raiyani and Roy Bayot
9. The Effect of Using Features Computed from Generated
Offline Images for Online Bangla Handwritten Character
Recognition ..................................................................................................121
Shibaprasad Sen, Ankan Bhattacharyya and Kaushik Roy
v
vi Contents
10. Handwritten Character Recognition for Palm-Leaf
Manuscripts .................................................................................................145
Papangkorn Inkeaw, Jeerayut Chaijaruwanich
and Jakramate Bootkrajang
Index .....................................................................................................................163
Preface
We are surrounded by huge volumes of data in various categories. In the con-
text of big data, automated and faster tools are in great demand. Often, data
exist on scanned documents and can be either machine-printed or hand-
written. These include, but are not limited to, letters, checks, payment slips,
income tax forms and business forms numbering in the billions. Manual
digitization is impossible, time consuming, expensive and vulnerable to
errors. Automation can solve such problems. Compared to machine-printed
documents, handwritten document processing can be more challenging
and, of course, opens a wide range of issues to consider. Writer identifica-
tion via handwriting can be one of the important applications. Further, we
have experienced that both machine-printed and handwritten texts can go
together in one document. Can we treat/process them separately? For this,
one needs to be able to extract meaningful information; it can be from either
machine-printed or handwritten texts. This can open up ideas of document
information extraction and understanding, such as information retrieval
and natural language processing.
Advances in document processing can only be possible by considering
appropriate machine learning algorithms that help build data-driven mod-
els or predictions. The process includes classification, clustering, regression,
association rules and many more elements. How can we forget deep learn-
ing-based approaches? In this book, advanced document processing tech-
niques using up-to-date and well-trained machine learning algorithms are
presented.
Chapter 1 discusses the role of AI for different document analysis prob-
lems, such as optical character recognition and web text categorization,
where real-world issues are considered.
Chapter 2 discusses multiple methods for optical character recognition
on handwritten isolated Bangla characters. It provides results on a database
of 89,000 plus handwritten characters in addition to two publicly available
benchmark databases, namely the ISI handwritten character database and
CMATERdb 3.1.2.
Chapter 3 presents a script identification study on multi-script artistic
documents where real-world problems are taken into consideration. The pri-
mary motivation behind this work is that OCR tools are script-dependent. In
the authors’ work, a semi-automated segmentation algorithm has been used
for character separation within words followed by a thinning procedure and
a structural feature; Gabor filters are used for feature extraction. The results
were reported using several different machine learning classifiers.
Chapter 4 discusses the use of extreme learning machines (ELMs), as
they suppress the pitfalls of neural networks like slow learning rates and
vii
viii Preface
parameter tuning. The authors provide an idea of how document analysis
can be done using an ELM.
Internet technology has brought about substantial changes to our day-to-
day lives. Nowadays, we need several digital means to automatically man-
age, filter, store and retrieve information from text documents. Automated
text document analysis and mining are becoming an essential part of com-
puter applications and thus various classification and clustering approaches
are required for carrying out these tasks. The classification of documents
needs to be performed in the training dataset, which is further used to train
the model classifier to classify the text documents into their respective text
categories or domains. Thus, text analysis becomes one of the major aspects
of text data mining. Under this scope, Chapters 5, 6 and 7 provide various
graph-based models for different web text classification problems.
Chapter 8 discusses two essential aspects of an author, namely author
aggression and author profile. The identification of aggression is classified
into three classes: overtly, covertly and non-aggressive. On the other hand,
profiling identifies the two properties of age and gender. The chapter also
explains various machine learning concepts incorporated into the deep neu-
ral network. It also provides insight into the previous work done on author
aggression and profiling.
Document image analysis problems broadly fall under two categories:
offline and online. In Chapter 9, the authors discuss online handwrit-
ten Bangla character recognition using features computed from generated
offline images. Interestingly, this chapter shows an integrated version/con-
cept of how offline and online domains can be merged to advance recogni-
tion performance.
Chapter 10 discusses handwritten character recognition for palm-leaf
manuscripts. Palm-leaf manuscripts are ancient documents primarily from
South and Southeast Asia. Over hundreds of years, the manuscripts have
become damaged. The chapter discusses how to transform the palm-leaf
images into machine-encoded texts so that the document can be edited,
retrieved, accessed and processed. The issue of applying optical handwrit-
ten character recognition for the said problem, with associated challenges, is
covered in detail.
We hope this book will help researchers in document analysis and machine
learning as different real-life problems are discussed with experimental out-
comes. Additionally, undergraduate or postgraduate scholars who wish to
carry out their research-based projects or theses on document analysis prob-
lems can get help from this book.
Sk Md Obaidullah
KC Santosh
Teresa Gonçalves
Nibaran Das
Kaushik Roy
Editors
Sk Md Obaidullah earned a PhD (Engg.) from
Jadavpur University, an MTech in Computer Science
and Application from the University of Calcutta and
a BE in Computer Science and Engineering from
Vidyasagar University in 2017, 2009 and 2004, respec-
tively. He was an Erasmus post-doctoral fellow funded
by the European Commission at the University of
Évora, Portugal, from November 2017 to September
2018. He has more than 11 years of professional experi-
ence including two years in industry and nine years
in academia, out of which five years were spent on research. Presently, he is
working as an Associate Professor in the Department of Computer Science
and Engineering, Aliah University, Kolkata. He has published more than 65
research articles in renowned journals and reputed national/international
conferences. He is an active researcher in the fields of document image pro-
cessing, medical image analysis, pattern recognition and machine learning.
KC Santosh (Senior Member, IEEE) is an Assistant
Professor and Graduate Program Director for the
Department of Computer Science at the University of
South Dakota (USD). Dr. Santosh also serves the School
of Computing and IT, Taylor's University, as a Visiting
Associate Professor. Before joining USD, Dr. Santosh
worked as a research fellow at the U.S. National Library
of Medicine (NLM), National Institutes of Health
(NIH). He worked as a postdoctoral research scientist
at the LORIA research centre, Université de Lorraine,
in direct collaboration with industrial partner ITESOFT in France. He also
worked as a research scientist at the INRIA Nancy Grand Est research cen-
tre, France, where, he completed his PhD in Computer Science. Before that,
he worked as a graduate research scholar at SIIT, Thammasat University,
Thailand. He has published more than 120 peer-reviewed research articles,
authored two books (Springer) and edited ten books, journal issues and
conference proceedings. Dr. Santosh serves as an associate editor for the
International Journal of Machine Learning & Cybernetics. Dr. Santosh demon-
strates expertise in artificial intelligence, machine learning, pattern recog-
nition, computer vision, image processing, data mining and big data with
various application domains, such as healthcare and medical imaging, docu-
ment information content exploitation, biometrics, forensics, speech/audio
analysis, satellite imaging, robotics and the Internet of Things.
ix