How to reuse translated material
At the II Machine Translation Conference
held at the UAB
, a European project was presented for the automated creation of translation memories: Bicrawler
. Using this portal, translation memories can be created automatically from websites with excellent results.
What is Bicrawler?
Bicrawler is one of the projects carried out by the consortium Abu-MaTran
, with the cooperation of academic institutions and the private sector. In October, during the II Machine Translation Conference organised by the Autonomous University of Barcelona, Gema Ramírez
from the company Prompsit
, which participates in the consortium Abu-MaTran, presented this free portal which can be used to create translation memories from multilingual websites.
The portal enables us to create memories in a combination of 12 languages, although the project hopes to make more languages available. The portal is designed to feed machine translation systems by creating translation memories from websites with reliable translations. These same memories can also be used as a reference for our own translations.
At present, if we want to use translations of company content for which there is no database, that is, a translation memory, we can "align" the documents. To do so, current computer assisted translation programmes include tools which, based on elements such as the format, the numbers, the punctuation or a terminological glossary, will provide us with a correspondence between the source text and the translation. These programmes are not able to produce a fully precise alignment and therefore the involvement of linguists is necessary, which can be a very expensive process depending on the tool used and the initial results, which may vary significantly based on the languages and the format.
The future of alignment
The novelty introduced by Bicrawler, when creating automatic translation memories without the need to review the alignment manually, is that it uses machine translation technology to predict the results and thus evaluate the quality of the aligned segments. In this case, the human revision of the aligned texts using alignment tools is substituted by an automatic revision based on dictionaries and machine translation. The results are spectacular!