Czytelnia / Technologie tłumaczeniowe

Another central notion is that of terminological database, or termbase:    

Termbase:  Short  form  of  Terminology  database.  A  termbase  is  the  collection  of  information  on  a  term  or  concept  in  a  structured,  electronically  readable  way  combined  with  a  terminology  management  system.  It  is  mostly  used  synonymously  with  termbank,  though  some  terminologists  distinguish  them.  If  they  are  distinguished,  terminology  databases  do  not  include  the  organizational  environment  but  termbanks  do.' (Trippel 1999) (cf. Galinski 1998). In this thesis the terms termbank and termbase will be used interchangeably.

Terminology management tools are part of a larger group of software tools referred  to  as  computer-assisted  translation  (CAT)  tools.  CAT  is  defined  as  `direct  translation  by  humans with the help of a computer interface which makes translational expertise accessible  through  "translation-intelligent"  software'.  (Neubert  1991:56).  In  other  words,  CAT  applications are a group of software tools assisting translators, where the human knowledge  and linguistic competence are the key factors, and it is the human translator who plays the  dominant  role  and  makes  the  final  decisions  concerning  terminology  and  phraseology  choices.    Modern CAT tools,  referred  to as workbenches, consist of  a  number  of modules  or  components,  terminology  management  systems  being  part  of  them.  The  module  which  is  considered the central one though, is the translation memory module.    

`There  are  different  TM  programs  currently  available  on  the  market,  but they share similar features, albeit with some differences in speed and data  management.  Normally,  the core  of  TM  is  the  memory,  a  complex  database  where  source text  sentences  are aligned  side by  side  with  the  corresponding  target  text  sentences.  The  ways  in  which  the  memory  can  be  accessed  and  managed vary from one TM program to the other, but the philosophy behind  the tool is basically the same: reusing previous work.'(Rico Pe'rez 2001).    In  a  nutshell,  TM  tools  play  the  role  of  a  perfect  memory  that  can  be  accessed  anytime  during  the  translation  process.  It  is  a  memory  that  never  fails  to  retrieve  the  requested information and prevents the translator from struggling with the same translation  problem  twice.  The  fact  that  translation  memory  stores  aligned  sentence  pairs  in  source  language  (SL)  and  target  language  (TL)  makes  the  tool  extremely  useful  in  translating repetitive  texts  e.g.  technical  manuals.  When  a  new  document  is  being  translated  in  a  workbench  environment,  the  program  automatically  searches  the  translation  memory  for  identical or  similar  segments, and  whenever a match is  returned (exact  or fuzzy) it will  be  displayed in a special pane or grid or directly in the space where the target segment should  be entered.   

There  are,  however,  technical  texts  that  are  very  dense  in  terms  of  specialist  terminology,  but  do  not  contain  as  much  as  two  identical  sentences.  In  this  case  the  terminology management component of a workbench comes in as the right solution (Benis  1998). Thanks to terminology management modules even in the case of non-repetitive texts  we  can  still  benefit  from  the  workbench  packages  in  terms  of  speed  and  quality  of  translation, even though translation memory is not applicable.     

Other components which are normally part of workbench applications are alignment  tools  (applications  used  for  building  translation  memories  from  the  corresponding  SL  and  TL documents), analysis modules performing word frequency and repeatability calculations,  sometimes also database and project maintenance modules.  

At  this  stage  it  is  necessary  to  draw  the  distinction  between  two  terms  which  are  frequently  confused,  i.e.  computer-assisted  translation  (CAT), also  referred  to  as machineassisted  human  translation  (MAHT),  and  machine  translation  (MT).  While  it  clearly  transpires  from  the  very  term  that  MAHT  is  the  type  of  translation  where  the  human  translator plays the crucial role (cf. Feder 2001:51, Neubert 1991:57) it should be noted that  `MT aims at assembling all the information necessary for translation in one program so that  a  text  can  be  translated  without  human  intervention'  (Craciunescu  et  al.  2004).  The  difference between MAHT and MT applications is also in the output quality. In the case of  MAHT tools, the translations are usually of publishing quality. The up-to-date MT systems,  on  the  other  hand,  deliver  translations  of  unacceptable  quality  or  requiring  much  postediting.  However,  the  advent  of  new  MT  systems  applying  neural  networks  and  artificial  intelligence  technology  is  only  a  matter  of  time  and  we  may  expect  the  quality  of  their  output to improve (Champollion: 2001).