Multi-lingual

1
Multi-lingual multi-institutional distant
learning

• Example of an international master programme in
  Computational Linguistics
• 14-16 November, Blaubeuren, Germany

Nikolai Vazov Sofia University
Kiril Simov LML - BAS BulTreeBank project
Petya Ossenova Sofia University BulTreeBank
project
2
General goals of the programme

• to put together linguistics (linguists) and
  computer technologies (computer scientists)
• to put together foreign and local expertise
• to promote international multi-lingual
  cooperation in order to develop multi-lingual
  language electronic resources

3
Programme participants (1-2 year)

• Two project managment partners
• French Ministry of Foreign Affairs
• French Cultural Institute in Sofia

• Two academic partners
• University of Sofia (2 departments)
• University of Paris IV - Sorbonne

4
Programme participants (3 year)

• Three project managment partners
• French Ministry of Foreign Affairs
• French Cultural Institute in Sofia
• Agence Universitaire de la Francophonie

• Six academic partners
• University of Sofia (3 departments)
• University of Paris IV - Sorbonne
• LML - Bulgarian Academy of Sciences
• University of Montréal (RALI OLST)
• University of Iasi (Romania)
• RACAI (Romania)

5
Organisation (educational activities)

• Foreign participants
• 1-3 intensive teaching sessions
• distant follow up between the sessions
• distant examination
• Local participants
• successive modules (1-3 weeks each) accompanied
  by web-based courses
• distant tutorship after the course - individual
  work with students (the format 8 students/15
  professors allows for it)
• on-line personal library for each student
  (articles to read and discuss with the other
  participants)
• distant examination

6
Organisation (research activities)

• Carried out as individual tasks with twofold
  impact
• development of personal skills in manipulating
  electronic text data (using CLaRK, Perl, MySQL,
  XML, HTML)
• integration of individual tasks into the main
  goal of the team - creation of mono- and
  multi-lingual electronic resources

7
Organisation (research activities)

• Examples
• writing tokenizers for French (solved in
  TreeTagger)
• sentence boundaries identification (not entirely
  handled by TreeTagger, but indispensable for
  parallel corpora)
• named entity recognition
• temporal expression extraction
• abbreviations identification
• parenthetic expressions identification
• concordances (Bulgarian French)

8
On-line ressources and tools
Developed by the team
Other available ressources

• CLaRK system
• Morphological dictionary for Bulgarian
• Large tagged corpus of Bulgarian
• Concordances (French, Bulgarian)
• Temporal expressions extractor (French)
• Large archive of bilingual (French-Bulgarian)
  texts

• Large tagged corpus FRANTEXT
• Large bilingual (French-English) aligned corpus
  Hansard
• Taggers (TreeTagger and LATL)
• Le Monde sur CD-ROM (with integrated search
  engine)

9
Future work

• New (better targeted) master  Electronic
  language resources 
• Goals of the master defined the other way around
  research needs determine the course content and
  not vice versa
• Envisaged product parallel French-Bulgarian
  corpus with named entity identification
• So far collection of parallel texts, development
  of proper names transcription module