CRBLP

1
CRBLPs (Center for Research on Bangla Language
Processing) Activities and Achievements on Bangla
Language Processing, January 2007

• Naushad UzZaman
• CRBLP, BRAC U, Bangladesh
• http//www.naushadzaman.com

2
CRBLPs Activities

• Center for Research on Bangla Language
  Processing, CRBLP working on Bangla Language
  Processing since 2004
• 11 Research staff (9 Computer Science background,
  2 linguistics background)
• Students working part-time, doing internship
• 13 Summer 2006 Interns and 7 former members
• Motivation of open source
• Academic
• Offered course on language processing (CSE 431
  Natural Language Processing, offered at Spring
  2006 and Spring 2007 in BRAC U)
• Thesis on NLP
• Summer Internship

3
CRBLP Members (Full Time Staff Members)

• Dr. Mumit Khan email website
• Head, CRBLP and Associate Professor, CSE
  Department
• Matin Saad Abdullah email?
• Program Manager, CRBLP and Senior Lecturer, CSE
  Department
• Naira Khan email
• Linguist, CRBLP and Lecturer, English and
  Humanities (On Leave)
• Zahurul Islam email website
• Research Programmer, CRBLP and Part-time Faculty
  Member, CSE
• Naushad UzZaman email website
• Research Programmer, CRBLP and Part-time Faculty
  Member, CSE
• Md. Abul Hasnat, Research Programmer email
  website
• S. M. Murtoza Habib, Research Programmer email
  website
• Firoj Alam, Research Programmer email website

4
CRBLP Members (Part-time and Interns)

• Part Time Staff Members
• Kamrul Hayder, Language Consultant
• M. Abdur Rahman, Research Assistant
• Maruf Muqtadir, Research Assistant
• Summer 2006 Research Interns
• Fahim Muhammad Hasan
• M. Hammad Ali
• Ayesha Binte Mosaddeque
• Nafid Haque
• Yeasir Arafat
• Nizam Uddin
• M. Abdur Rahman
• Fahim Tawfique Chowdhury
• Munirul Mansur
• Md. Jahangir Alam
• Annajiat Alim Rasel
• Munshi Asadullah
• Salman Zaman

5
Areas of Research

• Document Authoring
• Information Retrieval
• Optical Character Recognition
• Pronunciation Generator
• Speech Processing
• Morphology
• Parts of Speech Tagging
• Syntax
• And also few other small research projects

6
Document Authoring, BanglaPad

• The current version of the BanglaPad includes the
  following features
• 1. Platform independent. (Current version tested
  on Windows and Linux).
• 2. Edit Bangla and English text in the same
  document.
• 3. Rich text editing with pictures and tables.
• 4. Export document as HTML. (You can develop web
  contents in Bangla using this feature!)
• 5. Support character encoding including UTF8 and
  UTF16.
• 6. Bangla and English Spell checking. (Bangla
  spelling checker uses Puspa Speller)
• 7. Bangla and English Search and replace.
• 8. Printing formatted document.
• 9. Three different skins for the editor.
• 10. Built-in keyboard driver for easy Bangla
  typing. (No need to install a keyboard driver).
• 11. Customizable Key-Maps for Bangla.
• 12. Easy to use Installer for Windows

7
Spelling Checker in BanglaPad
8
Rich Text Editing in BanglaPad and exporting to
HTML
9
BanglaPad Download and Team Members

• Download http//sourceforge.net/project/showfiles
  .php?group_id158301package_id180246
• Developers
• Zahurul Islam August 2005 - December 2005
• Naushad UzZaman August 2005 - December 2005
• Abdur Rahman January 2006 - present
• Maruf Muqtadir January 2006 - present
• Advisors
• Matin Saad Abdullah August 2005 - present
• Mumit Khan August 2005 - present

10
English to Bangla Transliteration

• Type phonetically in English, you will get
  similar sounding dictionary word. Can be used for
  Bangla text input with English keyboard.
• Developed by Naushad UzZaman
• Relevant Publication
• 1. Naushad UzZaman, Arnab Zaheen and Mumit Khan,
  A Comprehensive Roman (English) to Bangla
  Transliteration Scheme, Proc. International
  Conference on Computer Processing on Bangla
  (ICCPB-2006), Dhaka, Bangladesh, 17 February,
  2006. ?
• 2. Naushad UzZaman, Phonetic Encoding for Bangla
  and its Application to Spelling Checker, Name
  Searching, Transliteration and Cross Language
  Information Retrieval, Undergraduate Thesis
  (Computer Science), BRAC University, May 2005.

11
Pata, English to Bangla Transliteration
12
Spelling Checker

• Bangla Speller Sandbox Bangla Phonetic Speller
  (Puspa). Gives suggestion for misspelling words
  based on similarities in pronunciation.
  Implemented based on Double Metaphone phonetic
  encoding
• Developed by Naushad UzZaman
• Download http//sourceforge.net/project/showfiles
  .php?group_id158301package_id180247

13
Publications on Spelling Checker

• 1. Naushad UzZaman and Mumit Khan, A Bangla
  Phonetic Encoding for Better Spelling
  Suggestions, Proc. 7th International Conference
  on Computer and Information Technology (ICCIT
  2004), Dhaka, Bangladesh, December 2004.
• 2. Naushad UzZaman and Mumit Khan, A Double
  Metaphone Encoding for Bangla and its Application
  in Spelling Checker, Proc. 2005 IEEE
  International Conference on Natural Language
  Processing and Knowledge Engineering, pp.
  705-710, Wuhan, China, October 30 - November 1,
  2005.
• 3. Naushad UzZaman and Mumit Khan, A
  Comprehensive Bangla Spelling Checker, Proc.
  International Conference on Computer Processing
  on Bangla (ICCPB-2006), Dhaka, Bangladesh, 17
  February, 2006.
• 4. Naushad UzZaman, Phonetic Encoding for Bangla
  and its Application to Spelling Checker, Name
  Searching, Transliteration and Cross Language
  Information Retrieval, Undergraduate Thesis
  (Computer Science), BRAC University, May 2005.
• 5. Munshi Asadullah, Md. Zahurul Islam, and Mumit
  Khan, Error-tolerant Finite-state Recognizer and
  String Pattern Similarity Based Spell-Checker for
  Bengali, to appear in the Proc. of International
  Conference on Natural Language Processing, ICON
  2007, January 2007.

14
Puspa Spelling Checker
15
Search Engine

• Bangla search engine based on open-source search
  engine Nutch.
• Developed by M. Hammad Ali and Nafid Haque
• Relevant Publications
• 1. Nafid Haque, Hammad Ali, Mumit Khan, and Matin
  Saad Abdullah, Infrastructure for Bangla
  Information retrieval in the context of ICT for
  Development, to appear in the Proc. of
  International Conference on Systems, Computing
  Sciences and Software Engineering (SCSS 06) of
  International Joint Conferences on Computer,
  Information, and Systems Sciences, and
  Engineering (CISSE 06), December 4 - 14, 2006.
• 2. M. Hammad Ali, Nafid Haque, A Decentralised
  Approach to Information Retrieval for a
  developing country like Bangladesh, Education
  Without Borders 2007, Abu Dhabi, February 25 -
  27, 2007.

16
Search Engine example
17
Optical Character Recognition

• BanglaOCR is the Optical Character Recognizer for
  Bangla Script. It takes scanned images of a
  printed page or document as input and converts
  them into editable Unicode text. BanglaOCR allows
  users to train the data set from any document and
  observe the recognition performance.
• BanglaOCR developed by Md. Abul Hasnat and S M
  Murtoza Habib.
• Download http//sourceforge.net/project/showfiles
  .php?group_id158301package_id215908
• Another OCR implemented using Kohonen Network,
  developed by Shoeb Shatil.
• Download http//sourceforge.net/project/showfiles
  .php?group_id158301package_id180249

18
OCR Status

• OCR Application
• Status Version 0.1, Release candidate 1
• Status of Different Segments of OCR
• Document skew correction
• Bangla document skew corrector based on Radon
  transform. Status Complete.
• Segmentation
• Bangla line segmentation. Status Complete
• Bangla word segmentation. Status Complete
• Bangla character segmentation. Status Work in
  progress. The large number of combinations
  (consonant clusters and the non-spacing marks)
  complicates this task. This is omnifont, so must
  work with any typeface.
• Character/Symbol recognition
• Neural net based recognizer Fairly complete for
  the basic alphabet and a subset of the consonant
  clusters. The non-spacing marks pose a
  significant challenge.
• Hidden Markov Model (HMM) based recognizer
  Status First demo available.
• -Post Processing for OCR
• Post processing spelling checker for OCR
  corrects spelling mistakes due to unsuccessful
  recognition. Status First demo available.

19
BanglaOCR
20
OCR Related Publications

• 1. Md. Abul Hasnat, S M Murtoza Habib and Mumit
  Khan, Segmentation free Bangla OCR using HMM
  Training and Recognition, to appear in the Proc.
  of 1st International Conference on Digital
  Communications and Computer Applications
  (DCCA2007), Irbid, Jordan, 2007.
• 2. A. M. Shoeb Shatil and Mumit Khan, Minimally
  Segmenting High Performance Bangla OCR using
  Kohonen Network, to appear in the Proc. of 9th
  International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.
• 3. S. M. Murtoza Habib, Nawsher Ahmed Noor and
  Mumit Khan, Skew correction of Bangla script
  using Radon Transform, to appear in the Proc. of
  9th International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

21
Automated Pronunciation Generator

• Pronunciation Generator Input any Bangla word,
  this application will give the pronunciation of
  that word in IPA (International Phonetic
  Alphabet).
• Demo available online at http//student.bu.ac.bd/
  7Eu02201011/g2pweb/g2p1.htm
• Source code available online at
  http//student.bu.ac.bd/7Eu02201011/g2pweb/
• Developed by, Ayesha Binte Mosaddeque
• Relevant Publication
• Ayesha Binte Mosaddeque, Naushad UzZaman and
  Mumit Khan, Rule based Automated Pronunciation
  Generator, to appear in the Proc. of 9th
  International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

22
Bangla Pronunciation Generator
23
Speech Processing

• Text-to-speech
• Voice for Festival.
• Status First demo available, Developed by Firoj
  Alam
• Automatic Speech Recognition
• Isolated Speech Recognition, Developed by A K M
  Mahmudul Hoque
• Continuous Speech Recognition. Status First demo
  available. Developed by Md. Abul Hasnat

24
Speech Related Publications

• 1. Firoj Alam and Promila Kanti Nath, Bangla Text
  to Speech using Festival, Undergraduate Thesis
  (Computer Science), BRAC University, May 2006.
  Supervisor Mumit Khan.
• 2. A K M Mahmudul Hoque, Bangla Speech
  Recognition, Undergraduate Thesis (Computer
  Science), BRAC University, May 2006. Supervisor
  Mumit Khan.
• 3. Firoj Alam, Promila Kanti Nath and Mumit Khan,
  Text To Speech for Bangla Language using
  Festival, to appear in the Proc. of 1st
  International Conference on Digital
  Communications and Computer Applications
  (DCCA2007), Irbid, Jordan, 2007.

25
Morphology

• Morphology The branch of grammar which studies
  the structure or forms of words.
• Work done on Bangla Morphology
• Generative verb morphology using two-level rules
• Basic concatanative noun morphology with features
• Software developed Jkimmo, A Multilingual
  Computational Morphology Framework for PC-KIMMO.
  Developed by Md. Zahurul Islam.
• Download Jkimmo http//sourceforge.net/project/sh
  owfiles.php?group_id158301package_id180248

26
Morphological Analyzer Jkimmo
27
Morphology Related Publications

• 1. Sajib Dasgupta and Mumit Khan, Morphological
  Parsing of Bangla Words using PC-KIMMO, Proc. 7th
  International Conference on Computer and
  Information Technology, Dhaka, Bangladesh,
  December, 2004.
• 2. Sajib Dasgupta and Mumit Khan, Feature
  Unification for Morphological Parsing in Bangla,
  Proc. 7th International Conference on Computer
  and Information Technology, Dhaka, Bangladesh,
  December, 2004.
• 3. Sajib Dasgupta, Dewan Shahriar Hossain Pavel,
  Asif Iqbal Sarkar, Naira Khan and Mumit Khan,
  Morphological Analysis of Inflecting Compound
  Words in Bangla, Proc. 8th International
  Conference on Computer Information Technology
  (ICCIT), Islamic University of Technology (IUT),
  Dhaka, Bangladesh, 2005.
• 4. Md. Zahurul Islam and Mumit Khan, JKimmo A
  Multilingual Computational Morphology Framework
  for PC-KIMMO, to appear in the Proc. of 9th
  International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

28
Bangla Parts of Speech (POS) Tagging

• This application tags words in a sentence with
  the parts of speech of that word. Implemented and
  compared HMM, n-gram and Transformation based
  Brills POS Tagging for Bangla, Hindi and Telegu
  on different sized corpus. For Bangla it was
  compared on different sized tagset too.
• Developed by Fahim Muhammad Hasan.
• Relevant Publications
• Fahim Muhammad Hasan, Naushad UzZaman and Mumit
  Khan, Comparison of different POS Tagging
  Techniques (n-gram, HMM and Brill's tagger) for
  Bangla, to appear in the Proc. of International
  Conference on Systems, Computing Sciences and
  Software Engineering (SCSS 06) of International
  Joint Conferences on Computer, Information, and
  Systems Sciences, and Engineering (CISSE 06),
  December 4 - 14, 2006.

29
POS Tagging example
30
Syntax

• Syntax the grammatical arrangement of words in
  sentences
• Bangla syntactic analysis using
• Lexical Functional Grammar (LFG) formalism
• Head-driven Phrase Structure Grammar (HPSG)
  formalism
• Work done by Naira Khan, Ayesha Binte Mosaddeque,
  M Hammad Ali and Nafid Haque.
• Relevant Publications
• 1. Md. Nasimul Haque and M. Khan, Parsing Bangla
  using LFG An Introduction, BRAC University
  Journal, Vol 2, No. 2, 2005.
• 2. Naira Khan and Mumit Khan, Developing a
  Computational Grammar for Bengali using the HPSG
  Formalism, to appear in the Proc. of 9th
  International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.
• 3. Ayesha Binte Mosaddeque, M. Hammad Ali and
  Nafid Haque, Design of Head-Driven Phrase
  Structure Grammer for Bangla, Undergraduate
  Thesis (Computer Science), BRAC University,
  December 2006. Supervisor Mumit Khan.

31
(No Transcript)
32

• Small Research Projects

33
Bangla Grammar Checker

• Implemented a statistical Bangla grammar checker
  based on n-gram analysis.
• Developed by Md. Jahangir Alam.
• Relevant Publications
• Md. Jahangir Alam, Naushad UzZaman and Mumit
  Khan, N-gram based Statistical Grammar Checker
  for Bangla and English, to appear in the Proc. of
  9th International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

34
Bangla Text Categorization

• Implemented Bangla Text categorization based on
  n-gram analysis. Trained on Prothom Alo newspaper
  corpus on 6 different categories.
• Developed by Munirul Mansur.
• Relevant Publications
• 1. Munirul Mansur, Naushad UzZaman and Mumit
  Khan, Analysis of N-gram based text
  categorization for Bangla in a newspaper corpus,
  to appear in the Proc. of 9th International
  Conference on Computer and Information Technology
  (ICCIT 2006), Dhaka, Bangladesh, December 2006.
• 2. Munirul Mansur, Analysis of n-gram based text
  categorization for Bangla in a newspaper corpus,
  Undergraduate Thesis (Computer Science), BRAC
  University, August 2006. Supervisor Mumit Khan.

35
Analysis of Prothom-Alo newspaper Corpus

• Frequency analysis of 1 year Prothom-Alo
  newspaper corpus.
• Relevant Publications
• 1. Yeasir Arafat, Analysis and Observations From
  a Bangla news corpus, Undergraduate Thesis
  (Computer Science), BRAC University, August 2006.
  Supervisor Mumit Khan.
• 2. Yeasir Arafat, Md. Zahurul Islam and Mumit
  Khan, Analysis and Observations From a Bangla
  news corpus, to appear in the Proc. of 9th
  International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

36
Language Modeling, forward and backward n-gram

• Investigating the prospect of backward n-gram
  compared to forward n-gram for Bangla.
• Relevant Publication
• Naira Khan, Md. Tarek Habib, Md. Jahangir Alam,
  Rajib Rahman, Naushad UzZaman and Mumit Khan,
  History (forward n-gram) or Future (backward
  n-gram)? Which model to consider for n-gram
  analysis in Bangla?, to appear in the Proc. of
  9th International Conference on Computer and
  Information Technology (ICCIT 2006), Dhaka,
  Bangladesh, December 2006.

37
Font Converter

• Converts different TTF fonts to Unicode encoding.
  Status Completed for Ullash, Prothoma, Bangsi
  Alpona fonts.
• Developed by Md. Zahurul Islam.
• Download http//sourceforge.net/project/showfiles
  .php?group_id158301package_id180250

38
Stemming

• Stemming Stemming is an algorithm developed to
  reduce a search query to its stem or root form,
  in other words, variations of particular words
  such as past tense and plural and singular usage
  are taken into account when performing a search,
  For example, applies, applying applied matches
  apply.
• Relevant Publications
• Md. Zahurul Islam, Md. Nizam Uddin and Mumit
  Khan, A Light Weight Stemmer for Bengali and Its
  Use in Spelling Checker, to appear in the Proc.
  of 1st International Conference on Digital
  Communications and Computer Applications
  (DCCA2007), Irbid, Jordan, 2007.

39
Text Summarization

• Text summarization is the technique which
  automatically creates an abstract or summary of a
  text. In this study we investigate what works
  have been done in this area and implement an
  extraction based text summarizer for Bangla
  language.
• Relevant work
• Md. Nizam Uddin, "A Study on Text Summarization
  Techniques and an Approach for Bangla Text
  Summarization", Independent Study, Computer
  Science, BRAC University, December 2006,
  Supervisor Md. Zahurul Islam, Mumit Khan

40
Language Resources

• Lexicon
• Wordlist of 160 thousands words with 1st step
  parts of speech tags.
• Corpus
• 1 year Prothom alo newspaper corpus
• Charjapad and Boru Chandi Dash er kabbo corpus
  (Edited by Md. Abdul Hai and Anwar Pasha)

41
CRBLP Publications

• 2004
• ICCIT 2004 (Bangladesh) 3 (Morphology, Spelling
  Checker)
• Total 3
• 2005
• IASTED CI 2005 (Canada) 1 (Name Searching)
• IEEE NLP KE 2005 (China) 1 (Spelling Checker)
• IEE Mobility 2005 (China) 1 (Text Input System
  for Mobile)
• ICCIT 2005 2 (Morphology, Compiler)
• BU Journal 1 (Morphological Parsing)
• Undergraduate Thesis 1 (Phonetic Encoding)
• Total 7

42
CRBLP Publication cont.

• 2006
• ICCPB 2006 (Bangladesh) 4 (Corpus, Lexicon,
  Spelling Checker, Transliteration)
• ICCIT 2006 (Bangladesh) 11 (HPSG, Corpus
  Analysis, Text Categorization, Pronunciation
  Generator, Backward n-gram, Grammar Checker, Skew
  Correction, Traveler Information System, OCR
  using Kohonen Network, Mobile Messaging,
  Morphology)
• CISSE 2006 (Online) 2 (comparison of POS
  tagging, Bangla Information Retrieval)
• Undergraduate Thesis 9 (Skew Correction, Mobile
  Messaging, Speech Recognition, OCR using Kohonen
  network, Text to Speech, Corpus Analysis, Text
  Categorization, POS Tagging, HPSG)
• Total 24
• 2007
• ICON 2007 (India) 1 (Spelling Checker)
• DCCA 2007 (Jordan) 5 (Stemming, OCR, Text to
  Speech, Semantics, wireless LAN)
• EWB 2007 (Abu Dhabi) 2 (Information Retrieval,
  Localization)
• Total 8
• Till January 2007

43
CRBLP website

• http//www.bracu.ac.bd/research/crblp/