Czytelnia / Technologie tłumaczeniowe

3.8 System's responses

The  system's  response  if  the  search  criteria  are  not  met  is  another  point  worth  considering.  Usually  terminology  management  tools  display  a  hitlist  of  near  matches,  or  a  `term not found' message. Some tools also log the unfound terms in a special file, as well as  keep  a  history  of  searches.  For  translators  working  under  time  pressure,  getting  the  right  term right away is extremely important. Usually, they cannot afford delays caused by using a wrong term found on the hitlist. Therefore, it is vital that the hitlist distinguishes the fuzzy  matches from exact matches. In most  terminology management tools fuzzy  match must be  activated manually, so the translator realizes that the hits displayed in the search results may  not be appropriate.    Another important question is whether the tool is able to recognize a misspelled term.  The logical assumption is that if the tool offers a fuzzy match search option, it should be able  to return a misspelled term. The problem is, however, that sometimes a simple typing error  may  alter  the  initial  letter  or  letters  of  the  term,  thus  causing  the  tool  to  return  absolutely  irrelevant  matches.  This  means  that  the  type  of  the  spelling  mistake  is  not  without  significance.    One more important issue is the tool's response to compound term searches. As has  been  already said,  the  tool  usually  displays  the  list  of  nearest  matches.  With  the  advent  of  hybrid systems, combining CAT with MT, some of the tools offer the so-called `assemble'  function, which causes the  tool  to  automatically create  its own  proposals for  rendering the  unfound compound term or phrase. Again, the user must be aware of the fact that the output  string  is  not  a  validated  term,  but  a  proposal  of  the  MT  module,  and  as  such  should  be  approached with a dose of mistrust and subject to verification.    For the users of inflectional languages it may also be crucial to determine whether a  tool returns the canonical form of a word if the search string contains an inflected form. The  tools discussed within this project only return forms entered in the termbase. This means that  in  order  to  receive  a  base  form  of  a  given  term  we  need  to  activate  fuzzy  search,  right  truncation or wildcard search.    For the users of English it is in turn important to know whether the tool recognizes  spelling variants e.g. colour vs. color; or compound spelling differences e.g. hyphenated vs.  non-hyphenated variants.   

3.9 Input of information

The input of information into a termbase has a number of aspects. One of the most  important ones is whether it is possible to format characters and paragraphs in any way. It is  essential in the case of languages where diacritics must be used. Unfortunately, not all fonts  offer diacritical characters. Hence, it may sometimes be necessary to change the source text  font to another for the target language.  

A convenient terminology management tool should also offer a possibility to easily  edit both new and existing entries. Some of the simplest editing functions, except for manual keying in the  data, is copying, pasting, dragging  and  dropping,  deleting, redoing, undoing,  inserting, and changing the layout, search and replace. These functions are offered by most  contemporary applications, including terminology management tools.

3.10  Terminology extraction

Terminology  extraction  can  be  defined  as  an  automatic  collection  of  terminology  from  a  mono-  or  bilingual  corpus  of  texts.  Such  functionality  can  highly  speed  up  the  completion of pre-translation tasks which include terminology work, since the terminology  lists created in the process of automatic extraction can be quickly imported into a termbase.  Unfortunately,  efficient  terminology  extraction  tools  usually  do  not  go  along  terminology  management  modules  or  workbench  packages.  One  exception  is  the  `lexicon'  function  in  De'ja` vu, which is a very simple, but efficient terminology extraction tool. The manufacturers  of the other tools discussed do provide terminology extraction tools, however they need to be  purchased separately.