The most common machine translation systems

News

The most common machine translation systems

Nowadays, there are numerous machine translation systems that are available. The aim of this section of the thesis is to describe the most frequently used translation engines.

  1. SYSTRAN

SYSTRAN is the machine translation system that is used by Yahoo and AltaVista’s Babel Fish. As to the origins of SYSTRAN technology, the program  started to be developed after the enormous success of the Georgetown experiment in 1968. From that point on, the company provided translation solutions for the United States Department of Defence and the European Commission. The system covers a wide range of languages, however it is mostly focused on the languages pairs of English/French and English/Russian. This is for the reson that in the course of the Cold War the company was concentrated on the above languages (Peters, 2008).

The methods that are used in SYSTRAN reflect the old-fashioned ways of machine translation which is also reflected in their recent BLEU scores which do not rank as highly as companies that use different methods in their approach. SYSTRAN achieved a BLEU score of 0.2523 for Chinese to English translation with an unrestrained data track, which gave the system them third place in this category (Peters, 2008).

  1. Google

Google used to employ SYSTRAN’s machine translation solutions technology in the past. However, recently the company make a decision to design their own machine translation system. The system developed by Google managed to obtain BLEU scores of 0.4557 and 0. 2999 for Arabic to English and Chinese to English translations, respectively.

Google implemented use CBMT approach with SMT as the method. In addition, the company frequently uses English as an interlingual language that is employed in order to reduce dictionary/data expenditures in these situations in which there is not enough bilingual corpora for unpopular language pairs. Therefore, in these situations in which the Google translator is used to convey the meaning of the source text that is written in Vietnamese into Polish, it will be translated from Vietnamese to English and subsequently from English to Polish. It should be emphasised that Google possesses enormous amounts of data and computing power that the company is able to use.