stat946w18/Unsupervised Machine Translation Using Monolingual Corpora Only
From statwiki
Revision as of 18:38, 17 February 2018 by Pa2forsy (talk | contribs) (Created page with "File:Under-Construction-Sign.gif == Introduction == Neural Machine translation systems must be trained on large corpora consisting of pairs of pre-translated sentences. T...")
Introduction
Neural Machine translation systems must be trained on large corpora consisting of pairs of pre-translated sentences. This paper proposes an unsupervised neural machine translation system, which can be trained without using any such parallel data.
Overview of Translation System
The unsupervised translation system has four components:
- An unsupervised word-vector alignment system
- An encoder
- A decoder
- A discriminator
Overview of Objective
The objective function is the sum of three terms:
- The de-noising auto-encoder loss
- The translation loss
- The adversarial loss