Thoughts on Machine Translation

Maosong Sun
2016-01-01
Abstract:After a systematic survey of the development history of machine translation(MT) from a macroscopic perspective, with particular emphasis on the main development path of underlying methodologies and core technologies in MT, this talk gives a general picture of the milestones that marked the key points of a long journey for both theoretical study and practical accomplishment for the past seven decades. The latest fruitful development achieved in MT shows that, the paradigm shift from the traditional linguistic rule-based approaches to the so-called empirical approach, based on increasingly available amounts of “raw data” in the form of massive collections of texts and their translations, compounded by the advancement of machine learning studies, has become the driving force leading to the breakthrough in MT. Based on the above observation and analysis, the potential of using the state-of-the-art MT methodolgy to solve other similar key applications is also demonstrated.
What problem does this paper attempt to address?