Translating and Maintaining the Integrity of the Text
Is it possible to develop a system for machine translation able to overcome the many limitations of the existing systems, limits that in fact are preventing any practical application?
I believe so, but we need to abandon the idea of an easy and quick solution obtained by some statistical magic formula. Instead we need semantic comprehension of the content, a large quantity of conceptual information, and a great deal of work for each language to be managed.
To realize how disappointing the state of the art is, we just need to test the two main systems representing the current approaches:
- statistical technology used be Google language tools;
- linguistic technology (without semantics), for example the translation software made available by PROMT.
Have you ever tried these systems? I will talk about them very soon, so you may want to test them a little, so that you can compare your impressions.