Publicaciones

Advanced search

Abstract

Álvaro Peris. Continuous spaces in Statistical Machine Translation. 2015. Seminar of the MT group

Continuous space models rely in a distributed representation of words: A real-valued, dense and low-dimensional representation. Neural networks are the natural way of modelling and dealing with this representation. Recently, and aided with deep learning techniques, new and encouraging approaches have been developed. In the field of SMT, these models can be combined with the existing ones in order to improve translations or perform solely: the so-called neural machine translation approach. In this talk, different approaches for applying continuous spaces in SMT are reviewed.