The purpose of this project is to develop an innovative computer-assisted system to ease the generation of high quality translations. In ordder to do that, we propose a novel interactive-predictive approach that places the human expert at the core of the translation process and integrates a statistical machine translation (SMT) system in a interactive editing environment.
The expert guarantees high quality translation, while the SMT system aims at increasing the productivity of the expert by providing automatic translation suffices. These translations can be accepted, amended o rejected by the expert. Interactivity allows the system to take advantage of the translation prefix validated by the expert in order to improve the precision of the next translation suffices offered by the system. Indeed, this project provides an excellent framework in which new adaptive learning techniques can be explored, being our objective to achieve that the system learns from the expert corrections.
The potencial impact of the project is notable, both from the scientific and commercial viewpoints. This projects focuses on a scientific problem that is currently present in many natural language processing systems: their inability to friendly adapt to unexpected circumstances and learn from interacting with expert users. The adaptive and on-line learning techniques that will be developed in this project will provide a significative progress in the state of the art. Furthemore, in the long-term, this progress will also help translation companies to keep pace with the increasing demand of high quality translation.