Abstract: In this project, we are planning to develop a bi-directional translation system which can work between any two language pairs. In our method, translation templates for a pair of language will be learned from a set of translation examples by using machine learning techniques. These learned translation templates will be used in the translation of other texts. Our proposed mechanism learns new translation templates from similarities and differences between two given translation examples. Given two translation examples, the similar parts of the sentences in the source language should match the similar parts of the sentences in the target language. Similarly, the difference parts should correspond to the respective parts in the translation examples. Our translation template learner algorithm will base on this basic heuristic. Some parts of this approach has been implemented and tested on a small training set and produced promising results for further investigation. For this reason, we want develop our all learning algorithms and test them on large data sets.

Keywords:  Example-Based Machine Translation, Machine Learning, Natural Language Processing.

Principal Investigator:   Ilyas Cicekli, Ph.D.
Investigator:  H. Altay Guvenir, Ph.D.
Investigator:  Zeynep Oz

Duration:  August 1997 - August 1999

Soponsor:   TUBITAK  (The Scientific and Technical Research Council of TURKEY)

Budget:  2,404,000,000 TL (~ 15,000 USD in August 1997)

Publications:

• Final Report (in Turkish)
• H. Altay Güvenir, and Ilyas Cicekli, Learning Translation Templates from Examples, Information Systems, Vol. 23, No. 6, (1998), pp. 353-363. ( postscript copy )
• Zeynep Oz, and Ilyas Cicekli, Ordering Translation Templates by Assigning Confidence Factors, in: Proceedings of AMTA'98-Conference of the Association for Machine Translation in the Americas, Lecture Notes in Computer Science  1529, Springer Verlag, October 1998, Langhorne, PA, USA, pp:51-61. ( postscript copy )
• Zeynep Orhan , Confidence Factor Assignment to Translation Templates, M.S. Thesis,  Bilkent University, September 1998. ( postscript copy  )