Japanese to english machine translation using neural networks

• Main Authors: Carandang, Jose Luis, Depano, Marie Anne, Rous, Giscard
• Format: text
• Language: English
• Published: Animo Repository 1997
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/16726
• Institution: De La Salle University

The thesis is a research in MT which involves the development of a Japanese-English MT Prototype System. The prototype system is based on the four phases of MT, namely: Pre-process, Analysis, Transfer, and Generation. As the various processes in each phase is handled by conventional approaches, the neural network approach will be implemented in the structural transfer processes. The research explores the possibility of applying the principles of neural networks in this particular task of translation. The neural network model takes in Japanese sentence structures as input and generates the appropriate English sentence structure as output. The implementation of the neural networks in this task will involve a study in recently developed connectionist representation techniques whose architectures are generally recursive to allow structure sensitive inputs to be represented into fixed length vectors.