Hiroshi NAKAGAWA

1

The Situation in Japan- A Personal View-

• Hiroshi NAKAGAWA
• Information Technology Center,
• University of Tokyo,Japan
• E-mail nakagawa_at_r.dl.itc.u-tokyo.ac.jp
• Postal 7-3-1 Hongo, Bunkyo, Tokyo, 113-0033,
  JAPAN

2
Structure of Academic Societies in JAPAN

• Basic data of NLP research activities in Japan
• At least 700 NLP researchers ( 700 is the number
  of the members of the Association of Natural
  Language Processing(ANLP) Japan)
• Several major universities and Labs in computer
  related companies (NTT, Fujitsu, NEC, Hitachi,
  Toshiba, .) and more.
• Several major governmental laboratories
  (Communication Research Lab., Electro-Technical
  Lab., National Institute of Informatics)

3
Academic Societies in Japan -1

• ANLPAssociation of NLP (700 members) the core
  of NLP research activities in Japan
• SIG NLP of Information Processing Society
  JapanIPSJ (more than 500 members mostly
  overlapped with ANLP members)
• SIG NLC of Institute of Electronics and
  Communication EngineeringIECE (also overlapped
  with ANLP)

4
Academic Societies in Japan -2

• Japanese Society of AI JSAI NLP is a not so big
  part
• SIG FI of IPSJ, SIG DD of IPSJ rather small SIG
• These are mainly IR people.
• The trend of the 90s was that NLP people were
  flowing into IR applications.

5
Academic Society in Japan -3

• Linguistics related societies
• Cognitive Linguistics Society (new)
• Social Linguistics Society (2 years old)
• Japan Cognitive Science Society NL of this
  societyJCSS is one of its main part but rather
  linguistics oriented research group.

6
Researchers really are

• Active researchers commit two, three academic
  societies above mentioned.
• Our worry is that very few linguists are
  interested in NLP.
• The reason is that they feel some kind of gaps or
  barriers between them and corpus based statistic
  oriented NLP approaches of these days.

7
Historical Perspectives

• The 80s Big projects with MT-related issues,
  supported by MITI-related agencies
• The 5th generation computer project
• Mu machine translation project
• EDR concept dictionary project
• CICC Translation project among Asian languages
• The 90s - Now Smaller projects with diverse
  issues including NLP as their part, supported by
  diverse funding agencies

8
From big project of the 80s towards sharply
focused the 90s
The 90s
9
Projects currently going on

• JSPS project NLP in IE and IR, Basic research
  of generic NLP technologies
• TIT-CRL-NLL project Speech and Language,
  Summarization of spontaneous speeches, Speech
  corpus collection
• RWC Corpus collection and annotation as one of
  the research activities
• ATR Speech Translation less restricted subject
  domains, adding new languages like Chinese,
  Example-based Paradigm

10

• KA Matching fund project Knowledge acquisition
  from comparable corpora
• Center for excellence on theoretical linguistics
  Language teaching as one of its application
  fields, Corpus collection
• GDA Text annotation in multimedia environments ,
  now extending to multimedia, multi-modal
  presentations
• --------(in preparation)---------
• Language and Action
• Usability Language in the network era

11

• Bottom-Up Initiatives
• IREX, NTCIR TREC,MUC-type workshop
• GSK LDC, ELRA-type institution
• Development Projects
• MT for National Patent Office MT in an
  integrated Information System
• JST MT for Abstracts, Example-based MT
• Automatic Caption System at NHK

12
TREC,MUC type bottom up activities in Japan

• IREX is a competition type workshop about
  Information retrieval and extraction, MUC style
  conference/ Evaluation, Training and test corpus
  collection
• NTCIR-1(1999) is also a competition type workshop
  about IR, CLIR, IE(named entity recognition), and
  ATR. Up to 40 groups were participated.
• NTCIR-2(2000) takes IR, CLIR, Automatic
  Summarization as its tasks ? i.e. about 20
  groups have been participated to CLIR task so far.

13
Real effects of the 80s big projects

• The 80s big projects have brought up core NLP
  researchers of the 90s.
• Funding agencies started to think Big fund is
  inefficient. They prefer small, clear target
  project.
• NLP researchers inclined to do realistic
  applications. (Good or bad?)
• This NLP research trend is timely fitting well
  todays IT revolution.
• For instance, NLP for cell-phone like gears is
  the promising.

14
International Co-operation The current state
Co-operation based on individual projects
JSPS project DFKI, U-Penn, Stanford Univ.,
KAIST UMIST/Salford,
etc ATR C-Star Consortium
Co-operation based on private companies
Co-operation in broader communities, supported by
Government agencies No since CICC, MT among
Asian languages
Human Genome, Physics, Space Science, Brain
Science
Language-based technology Highly sensitive
to the priority as a nation Too closely
related with industrial interests
15
Needs for Co-operation

• Validation of research results
• Language universal vs. Language specific
• NE, IR, Learning methods, grammar formalisms
• Resource gathering
• Simply many languages to treat
• Integration of resources
• Ontology for multi-lingual applications

Independent funding
Sponsoring/Independent funding
Standardization
Co-ordinated/Joint funding
16
From the 80s big project towards the sharply
focused 90s
Targets
Sponsoring/Independent Funding Standardization
Independent Funding
The 90s
Coordinated Funding