Walter von Hahn * Machine Translation

3. Types of Use¹

Looking more closely to the functional types of translations, Carbonell distinguishes two categories: "dissemination" and "assimilation". Only for dissemination the quality requirements are high, because:

1. in most cases the reader is unknown to the writer, and
2. the text contains exactly the information to be shared with the reader,

whereas in assimilation the amount of information to be extracted is dependent on the specific interest of the reader, which may vary between "to know what it is about" and "what exactly does the author mean?".

Technological Types and Dimensions

Another view to today's translation technology is the overall functionality and the processing type, which can be categorized along three dimensions:

• Functionality (What does the system do for its user?)
  • Machine aided translation
  • Machine translation
  • Machine interpreting
  
  
• Processing depth (Which steps between source and target?)
  • Shallow matching
  • Transfer
  • Interlingua
  • Variable depth
  
  
• Processing Paradigm (Which basic principles?)
  • knowledge-based
  • rule-based
  • stochastic
  • connectionist
  • example-based
  • semantic parsing

4. State of the Art - Research²

The following statements characterize the (very heterogeneous) state of research in MT. However, they emphasize the changes in interest and paradigms in the MT community:

• rule based paradigm is less abstract than that of the "indirect" models,
• syntactic analysis is restricted to surface constituency/dependency relations (roles and cases); single mono-stratal representation (unification/constraint-based analysis),
• semantic analysis is limited mainly to identification of sentence and clause roles (agents, patients, etc.),
• lexical information derived primarily from standard dictionary sources ("crude" syntactic categories and semantic features),
• lexical/ structural transfer rules (constraint-based) operating on shallow representations,
• example translations (aligned bilingual corpora),
• statistical data about lexical collocations and vocabulary frequencies (monolingual),
• probabilities of lexical transfer (bilingual),
• domain specific knowledge bases (both linguistic and subject knowledge),
• feedback ("learning") for grammar/lexicon improvement (connectionist?),
• greater emphasis on discourse and text stylistic aspects,
• integration into documentation processing and publishing systems.

122