GOALS
The main objective of METIS is to assess the possibility of developing a Statistical Machine Translation system which relies on annotated monolingual corpora and returns reasonable translations of free text.
This main goal can be broken up into the following sub-goals:
- Specify the exact requirements in terms of translation accurracy in order to design and evaluate the system according to a coherent plan
- Develop the algorithms which form the core of the system and are language independent
- Develop/adapt the necessary bilingual resources and offer some general methodological instructions for this task
- Evaluate the performance of the system with a view to its further development and integration to real-world applications
This project aims to contribute to the EU scientific and social objectives as well as the EU's economic development by investigating the possibility of improving the machine translation component of applications which deal with multilinguality issues.
The objective of METIS is to assess the possibility to obtain free text translations of reasonably high linguistic quality from large annotated monolingual corpora with pattern-matching techniques.
The existing monolingual resources are corpora, taggers and lemmatisers for Greek and Dutch as well as bilingual (Dutch-English, Greek-English) lexica.
By the end of the project:
- The bilingual resources will have been adapted to meet the project's needs.
- The algorithms which will perform the pattern-matching based translation will have been developed.
- Two language pairs (Dutch-English, Greek-English) will have been developed and evaluated.