Final Evaluation Review 12 October 2004 Luxembourg

1
Final Evaluation Review12 October
2004Luxembourg

• EDC-22071 KOM2002/27938
• Terminological Workbench and Customization
  Dictionaries
• Elsa Sklavounou, Ph. D.
• Linguist, Co-funded Projects Technical
  Coordinator
• SYSTRAN

2
Mental Health Terminology State-of-the art

• Lack of mental health existent glossaries in all
  KOM2002 language pairs.
• The basic resources used for the construction of
  the Web4Health dictionary were
• the SYSTRAN medical domain dictionary
• the list of all the Not Found Words were
  extracted from the produced EN FAQs for all the
  KOM2002 language pairs translated by content
  partners and SYSTRAN
• the ICD10 terminology translated in German (800
  entries), Greek (2200 entries) and Italian (2200
  entries) and
• the general medical dictionary
  http//allserv.rug.ac.be/rvdstich/eugloss/multi00
  1.html (3000 entries) created by Heymans
  Institute of Pharmacology, University of Gent and
  Mercator College, Department of Applied
  Linguistics Gent, Belgium December, 1995 and
  contracted by the Commission of the European
  Communities ETD/93/A02600/MI/10.

3
KOM2002Customization MethodologyStep 1

• SYSTRAN Medical domain dictionary
• The systems medical domain dictionary has been
  activated in priority so that homographs and
  polysemous entries are translated according to
  the specialized medical lexical resources. The
  systems medical dictionary has 2.000 entries.
• ICD-10 terminology
• The University of Ioannina provided us with the
  translation of ICD-10 index (2300 entries) into
  Greek. Terms have been sent to content partners
  for translation. ABIT provided the translation
  for 2300 mental health terms in Italian and FH
  NON 812 mental health terms into German.
• General Medical Dictionary
• 2.000 technical entries have been collected from
  the general medical dictionary according to the
  coverage on Web4health content.

4
KOM2002Customization MethodologyStep 2

•  Not Found Words Extraction
• There are 2.500 Not Found Words (most of them
  Proper Nouns and medical terms) extracted from
  the Web4health content.
• The list has been sent to the partners for
  translation of the medical terms in order to get
  the expert and medically correct translation. The
  enit and the ende dictionaries have been
  constructed thanks to ABIT and FH NON feed back
  (500 and 1300 entries respectively).
• 30.000 words have been translated from English
  into Swedish

5
KOM2002Terminological Workbench

•  
• The external mental health dictionaries were
  compiled via the Web4health terminological
  workbench and loaded to the SL API KOM2002
  account so that they are called during the
  translation processes in all language pairs.
• During the compilation of the dictionary entries,
  Intuitive Coding attributes to the bilingual
  entries all the elementary morpho-syntactic
  properties. Semantic or advanced information can
  be attributed to the entries by the expert coding
  and dictionary lookup to SYSTRAN resources.

6
KOM2002User Dictionary Compilation

• MULTI
• EN DE
• therapy (noun) Therapie (noun)
• sufferer (noun) Betroffener (noun)
• "PhD" (acronym) "Dr." (acronym)
• nursing home (noun)Schwesternhaus (noun)
• disorder (noun) Störung (noun)
• diet (noun) Diät (noun)

• AUTHOR_ BROWSE_
• CSRC_0 therapytherapyNN51PROGENMSCTABS0
  .999
• CSRC_100
• sufferersuffererNN11HUMANSREQMODHUCT0.9
  99
• CSRC_3 PhDPhDACRACR0.999
• CSRC_101 CSRC_5 nursing homenursinghomeNA15N1
  1LOCATNCONENCCONNCTCONABS0.998
• CSRC_43 disorderdisorderNN11PROPTYMSABS0
  .999
• CSRC_8 dietdietNN11CTABS0.999

7
Terminological Workbench MT Customization
8
Terminological Workbench Dictionary Management
9
Terminological Workbench Dictionary Update