Towards a Quality Improvement in Machine Translation: Modelling Discourse Structure and Including Discourse Development in the Determination of Translation Equivalents

Recent results in the field of machine translation research and development have shown that there will be much progress in future. The results have even encouraged researchers to see indications "that the time may finally be coming to strive once again for the promise of true automated translation" [Carbonell/Tomita 87:69]. With respect to a more general applicability of the methods involved one of the most crucial research requirements is the ability to cope with the aspect of textuality in translation. The requirements stated include: a text theoretical foundation of textual operations (such as text analysis, text generation, text translation) (cf. e.g. [Tsujii 86,88], [Hauenschild 88], [Nagao 89]), text analysis/text understanding components (cf. e.g. [Bar-Hillel 60], [ALPAC 66] [Carbonell/Cullingford/Gershman 81], [Tsujii 86,88], [Carbonell/Tomita 87], [Nagao 87,89], [Nirenburg/Raskin/Tucker 87], text generation components (cf. e.g. [Tsujii 86], [Bateman 90]).