Chapter 7

Statistical Methods for Machine Translation 1

7.1. Introduction

7.1.1. Machine translation in the age of the Internet

Machine Translation (MT) is a long established research area, which has, over the years, fostered the development of many technologies now widely accessible both to professional translators and the general public. Research on machine translation can be traced back to the early 1950s and was initially strongly influenced by the American mathematician Warren Weaver. Weaver compared a foreign language with a (secret) code to be deciphered, at a time when computers had already proven useful for breaking such codes. The history of the development of machine translation technology has not been linear and has given rise to many hopes and delusions (see, for instance, the writings of John Hutchins on the history of MT (e.g. [HUT 92, HUT 01, HUT 03]). The slow pace of the progresses of MT systems and the poor quality of automatic translations have often been mocked. Nonetheless, improvements have been steady and genuine and the actual situation of the domain, before the great “statistical shift”, can be roughly described as follows.

Several professional MT systems were on the market, delivering automatic translations for multinational industries and multilingual institutions such as the European Union and the Canadian government. These systems, an example of which is the software produced by Systran, relied primarily on manually developed translation rules, integrating ...

Get Textual Information Access: Statistical Models now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.