ASAR 2017


The 1st International Workshop on Arabic Script Analysis and Recognition (ASAR 2017), was hosted by LORIA laboratory (University Lorraine, France), in collaboration with the REGIM-Lab. (University of Sfax, Tunisia), and will be held in Nancy (France) from April 3-5, 2017.
ASAR 2017 was technically co-sponsored by IEEE Region 8, IEEE France section, IEEE Tunisia section, and IAPR.

The first ASAR 2017 workshop provided an excellent opportunity for researchers and practitioners at all levels of experience to meet colleagues and to share new ideas and knowledge about Arabic script document analysis and recognition methods. The workshop enjoys strong participation from researchers in both industry and academia.



In Memory of Professor Adnan Amin

Adnan_ASARThis first edition of ASAR 2017 was held as a tribute to Professor Adnan Amin (1951-2005). Adnan presented his doctorate (Doctorat d’Etat) in Nancy University in 1985 and spent a few years in this University as an Associate Lecturer. After several years in France and Kuwait, he joined in the School of Computer Science of the University of New South Wales in 1991. Adnan Admin contributed to develop the field of Arabic writing recognition. He was a pioneer, developing the first recognition systems and encouraging young researchers. He also was a man of peace and dialog between cultures. We all remember his kindness and friendliness.


Monday 3 April 2017

  • 8:30am: Registration
  • 9:30am: Opening, A. Belaïd & A. Alimi
  • 10am: Keynote 1: Gernot A. Fink, Statistical Methods for Arabic Handwriting Recognition (Chair: Adel M. Alimi)
  • 11am: Break
  • 11:30am: Oral Session 1 : Text Detection and Recognition in Videos and Scenes (Chair: Abdel Belaïd)
    • #11 Mohit Jain, Minesh Mathew and C.V. Jawahar. Unconstrained Scene Text and Video Text Recognition for Arabic Script
    • #16 Hashem Ghaleb, P. Nagabhushan and Umapada Pal. Segmentation of Offline Handwritten Arabic Text  (will be presented by Abdel Belaid)
    • #66 Seiya Iwata, Wataru Ohyama, Tetsushi Wakabayashi and Fumitaka Kimura. Recognition and Connection of Moving Captions in Arabic TV News
    • #61 Maroua Tounsi, Ikram Moalla and Adel M. Alimi. ARASTI: A Database for Arabic Scene Text Recognition
  • 1pm: Lunch
  • 2pm: Oral Session 2 : Segmentation (Chair: Laurence Likforman-Sulem)
    • #13 Ahmad Montaser Awal and Abdel Belaid. Neighborhood Label Extension for Handwritten/Printed Text Separation in Arabic Documents
    • #29 Hanadi Mohammed and Sumaya Al-Maadeed. Arabic Handwriting Recognition using Sequential Minimal Optimization (will be presented by Kalthouom Adam)
    • #10 Zineb Hadjadj* and Abdelkrim Meziane. Binarization of Document Images with Various Object Sizes
    • #39 Samia Snoussi and Yosra Wahabi. Arabic Document Segmentation on a smartphone towards Big Data HAJJ rules extraction (will be presented by Adel M. Alimi)
    • #67 Riaz Ahmad, Muhammad Zeshan Afzal, Sheikh Faisal Rashid, Marcus Liwicki and Andreas Dengel. Text-Line Segmentation of Large Titles and Headings in Arabic Like Script
  • 3:40pm: Break
  • 4:10pm: Oral Session 3: Online handwriting and health services (Chair: Christopher Kermorvant)
    • #44 Abdulaziz Alayba, Vasile Palade, Matthew England and Rahat Iqbal. Arabic Language Sentiment Analysis on Health Services
    • #21 Catherine Taleb, Laurence Likforman-Sulem, Maha Khachab and Chafic Mokbel. Feature Selection for an Improved Parkinson’s Disease Identification Based on Handwriting
    • #26 Zouhaira Noubigh and Monji Kherallah. A survey on Handwriting Recognition based on the Trajectory Recovery technique (will be presented by Adel Alimi)
    • #32 Afef Kacem and Abdel Belaid. Impact of Features and Classifiers Combinations on the Performances of Arabic Recognition Systems
    • #26 Zouhaira Noubigh and Monji Kherallah. A survey on Handwriting Recognition based on the Trajectory Recovery technique
  • 6pm: Welcome Reception (LORIA Hall)

Tuesday 4 April 2017

  • 9:20am: Oral Session 4 : Datasets (Chair: Habib M. Kammoun)
    • #6 Majeed Kassis, Reem Alaasam, Alaa Abdalhaleem, Ahmad Droby and Jihad El-Sana. VML-HD: The Historical Arabic Documents Dataset for Recognition Systems
    • #25 Alaa Abdalhaleem, Ahmad Droby, Abed Asi, Jihad El-Sana, Majeed Kassis and Reem Alaasam. WAHD: A dataset for Writer Identification of Arabic Historical Documents
    • #56 Mohamed Meddeb, Hichem Karray and Adel Alimi. Building and Analysing Emotion Corpus of The Arabic Speech
  • 10:20am: Break
  • 10:50am: Discussion group and Synthesis (Chair: Jihad El Sana)
  • 1pm: Lunch
  • 2pm: Oral Session 5: Feature-based Recognition (Chair: Gernot A. Fink)
    • #12 Mohammed Faouzi Benzeghiba. Arabic Word Decomposition Techniques for Offline Arabic Text Recognition
    • #27 Edgard Chammas, Chafic Mokbel and Laurence Likforman-Sulem. Stroke width exploitation to improve automatic recognition of arabic handwritten texts
    • #51 Qurat Ul Ain Akram and Sarmad Hussain. Ligature-based Font Size Independent OCR for Noori Nastalique Writing Style
    • #4 Liren Chen, Ruijie Yan, Liangrui Peng, Akio Furuhata and Xiaoqing Ding. Multi-layer Recurrent Neural Network based Offline Arabic Handwriting Recognition
  • 3:20pm: Break
  • 3:50pm: Oral Session 6: Historical documents (Chair: Abdel Belaïd)
    • #2 Andrei Boiarov, Alexander Senov and Alexander Knysh. Arabic Manuscript Author Verification Using Deep Convolutional Networks
    • #23 Tayyeba Faisal and Somaya Almaadeed. Enabling Indexing and Retrieval of Historical Arabic Manuscripts through Template Matching Based Word Spotting (will be presented by Kalthouom Adam)
    • #35 Kalthouom Adam and Sumaya Al-Maadeed. Character-based classification of Arabic scripts style in ancient Arabic manuscript
    • #47 Stephen Rawls, Huaigu Cao, Ekraam Sabir and Prem Natarajan. Combining Deep Learning and Language Modeling for Segmentation-Free OCR From Raw Pixels
    • #49 Reem Alaasam, Majeed Kassis, Jihad El-Sana, Alaa Abdalhaleem, Ahmad Droby and Berat Kurar Barakat. Experiment Study on Utilizing Convolutional Neural Networks to Recognize Historical Arabic Handwritten Text.
  • 7:45: Gala Dinner: “La Bergamote” cruise restaurant

Wednesday 5 April 2017

  • 9:30am: Keynote 2: Christopher Kermorvant, Handwriting recognition: is it a solved problem?  (Chair: Abdel Belaïd)
  • 10:30am: Break
  • 11am: Oral Session 7: Text and Style recognition (Chair: Jihad El Sana)
    • #33 Kawther Khazri, Afef Kacem and Abdel Belaid. Arabic/Latin and Handwritten/Machine-printed Formula Classification and Recognition
    • #63 Hanen Akouaydi, Slim Abdelhedi, Sourour Njah, Mourad Zaied and Adel M. Alimi. Decision Trees Based on Perceptual Codes for On-Line Arabic Character Recognition
    • #64 Ido Kissos and Nachum Dershowitz. Image and Text Correction Using Language Models
    • #36 Raid Saabni. Boosting Feature Based Classifiers for Writer Identification.
    • #62 Makki Maliki, Naseer Al-Jawad and Sabah Jassim. Off Line Writer Identification for Arabic language: Analysis and Classification techniques using Subwords Features
    • #37 Nabil Ghanmi and Ahmad Montaser Awal. Dynamic bayesian networks for handwritten Arabic words recognition (will be presented by Adel M. Alimi)
  • 1pm: Lunch
  • 2pm: Awards + Closing

Keynote Speakers:

Christopher Kermorvant

Founder of Teklia, France

Title: Handwriting recognition: is it a solved problem?

Abstract: Even if it was one of the first problem tackled by artificial intelligence, Handwriting recognition has not received the same attention as speech recognition or machine translation, both from the researchers and the public. In the shadow of already existing products for modern printed text recognition (OCR), the status of handwriting recognition is not clear for people outside the research community: is it a solved problem or not?

To answer this question, I will review the different levels of evaluations for handwriting recognition technology in the light of international evaluations. I will describe the different parts of a complete text recognition system and present evaluations of complete systems in applications. Finally, I will discuss some aspects of the possible future for handwriting and handwriting recognition.

Biography: Christopher Kermorvant graduated in Computer Engineering from Ecole Nationale Supérieure d’Informatique pour l’Industrie et l’Entreprise (France), holds a M.Sc. in advanced computer science from the University of Manchester (UK) and a Ph.D. in computer science from the Université de Saint-Etienne (France). He was post-doctoral researcher at University of Montreal with Yoshua Bengio. He started working the field of Machine Learning by developing speech recognition systems at IDIAP (Switzerland). He joined the French company A2iA in 2005 to develop automatic classification of scanned documents. From 2007 to 2014, he led the A2iA research team on handwriting recognition and participated to many collaborative projects with French and European laboratories. With his team, he developed handwriting recognition systems based on deep neural networks that ranked at the top of many international evaluations in French, English and Arabic handwriting recognition (OpenHaRT11, OpenHaRT2003, Maurdor). These systems are now integrated in A2iA products. Since 2015, he works as an independent expert in Machine Learning and Document Recognition.

Gernot A. Fink

TU Dortmund University, Germany

Head of Pattern Recognition in Embedded Systems Group

Title: Statistical Methods for Arabic Handwriting Recognition

Abstract: In the early days of optical character recognition research, systems processed
script in two distinct steps: First, the script would be segmented into characters and, second, these would be mapped to recognition hypotheses by some classifier. This approach works fine for tasks where the segmentation problem can be solved easily as, e.g., the recognition of machine printed text. For handwritten text, however, where segmentation becomes an issue or for any type of cursive writing it is desirable to optimize segmentation and recognition in a single integrated framework.

This became possible with the introduction of the statistical paradigm to the field of optical character recognition in the 1990ies. Then extremely successful recognizers could be built based on the concept of Markovian models. This technique was first applied to the recognition of handwriting in Roman script. Later, it was also applied to the recognition of machine printed Arabic texts as Arabic script is inherently cursive in nature and character segmentation is non-trivial. Subsequently, the more challenging problem of recognizing Arabic handwriting was also addressed successfully by systems based on the typical combination of hidden Markov models (HMMs) for modeling script appearance and statistical language models for capturing long-term contextual constraints. Therefore, until just recently, most successful systems for Arabic handwriting recognition were based on these techniques.

In this presentation, I will first introduce the basic concepts of recognizers based on Markovian models. Then I will briefly describe the historical development of statistical recognizers in the context of handwriting recognition with special focus on Arabic script. Afterwards, I will highlight some important structural aspects that set recognizers for Arabic scipt apart from their counterparts applied to Roman handwriting. Finally, I will give a short outlook on future trends in the field that, due to the overwhelming success of modern neural network architectures, will either go towards hybrid systems or even pure neural recognizers featuring deep architectures.

Biography: Gernot A. Fink received the diploma in computer science from the University of Erlangen-Nuremberg, Erlangen, Germany, in 1991 and the Ph.D. degree (Dr.-Ing.) also in computer science from Bielefeld University, Germany, in 1995. In 2002 he received the venia legendi (Habilitation) in Applied Computer Science from Bielefeld University.

From 1991 to 2005 he was with the Applied Computer Science Group at the Faculty of Technology of Bielefeld University. Since 2005 he is professor for Pattern Recognition in Embedded Systems within the Department of Computer Science at TU Dortmund University, Dortmund, Germany.

His research interests lie in the development and application of pattern recognition methods in the fields of man machine interaction, multimodal machine perception including speech and image processing, statistical pattern recognition, and handwriting recognition.

Dr. Fink has published extensively on the use of Markov-model based techniques for pattern recognition problems. He is the author of a textbook on Markov Models for Pattern Recognition and co-author of a survey article and a specialized monograph on the application of markov Models for handwriting recognition. He has been working on various problems in the fields of handwriting recognition and document analysis. Today, his team at TU Dortmund University is among the leading research groups in the field of word spotting and just recently, at ICFHR 2016, won both a track within the Word Spotting Competition and the Best Paper Award with a method based on deep learning, the so-called PHOCNet.


General Chairs
Abdel Belaïd (University of Lorraine, France)
Adel M. Alimi (University of Sfax, Tunisia)

Program Chairs
Laurence Likforman-Sulem (Telecom Paris Tech, France)
Jihad El-Sana (Ben-Gurion Univ. of the Negev, Israel)

Publicity Chairs
Haikal El Abded (GIZ, lebanon)
Volker Märgner (IFN, Braunschweig, Germany)
Rolf Ingold (Univ. Fribourg, Swizerland)

Organization Chairs
Habib M. Kammoun (University of Sfax, Tunisia)
Romain Karpinski (LORIA, Nancy, France)

Technical Program Committee
Ifran Ahmad (KFUPM Dhahran, Saudi Arabia)
Alireza Alaei (Griffith University, Australia)
Adel M. Alimi (REGIM-Lab., Tunisia)
Atallah AL-Shatnawi (Al al-Byte University, Jordan)
Somaya Al-Maadeed (Univ. Qatar, Qatar)
Abdel Belaïd (LORIA, France)
Saad Bin Ahmed (King Saud Bin Abdulaziz Univ., Saudi Arabia)
Mohamed Ben Halima (Univ. Sfax, Tunisia)
Jihad El-Sana (Ben-Gurion Univ. of the Negev, Israel)
Ashraf El-Sisi (Menofia Univ., Egypt)
Abdel Ennaji (LITIS – Univ. Rouen, France)
Najwa Essoukri Ben Amara (Univ. Sousse, Tunisia)
Saqib Bukhari (DFKI, Germany)
Youcef Chibani (Univ. STBH, Algeria)
Fadoua Drira (Univ. Sfax, Tunisia)
Djamel Gaceb (LIRIS-INSA de Lyon, France)
Nabil Ghanmi (LORIA, Nancy, France)
Venu Govindaraju (Univ. Buffalo, USA)
Tarek M. Hamdani (Taibah Univ., Saudi Arabia)
Afef Kacem (ESSTT, Tunisia)
Slim Kanoun (Univ. Sfax, Tunisia)
Monji Kherallah (Univ. Sfax, Tunisia)
Laurence Likforman-Sulem (Telecom Paris Tech, France)
Samia Maddouri (FCIT- Jeddah Univ., Saudi Arabia)
Driss Mammass (Univ. Ibn Zohr, DZ)
Ikram Moalla (Al-Baha Univ., Saudi Arabia)
Khader Mohammad (Birzeit Univ., Palestine)
Chafic Mokbel (Univ. Balamand, Lebanon)
Jean-Marc Ogier (Univ. La Rochelle, France)
Umapada Pal (ISI, Kolkata, India)
Muhammad Imran Razzak (King Saud bin Abdulaziz Univ., Saudi Arabia)
Raid Saabni (Yafo College, Israel)
Faisal Shafait (Univ. W. Australia in Perth, Australia)