The present paper addresses MT of asymmetrical linguistic markers, in particular zero possessives. English <-> Russian MT was chosen as an example; however, obtained results can be applied to other language pairs (English – German / Spanish/Norwegian etc.).Overt pronouns are required to mark possessive relations in English. On the contrary, in Russian implicit possessives are regularly used, thus making it very important to analyze them properly, not only for MT but also for other NLP tasks such as NER, Fact extraction, etc. However, concerning modern NLP systems the task remains practically unsolved. The paper examines how modern English <-> Russian MT systems process implicit possessives and explores main problems that exist concerning the issue. As no SB approach can process IP constructions properly, linguistic rules need to be developed for their analysis and synthesis;the main properties of IPs are analyzed to that end. Finally, several rules to apply to RB or model-based MT are introduced that help to increase translation accuracy. The present research is based on ABBYY Compreno ? multilanguage NLP technologies that include MT module.
展开▼
