The Global Watchtower blog contains a post about recent developments in rule-based MT.
For a brief overview of rules-based MT as opposed to its competing technology, statistical MT, about the differences between the two approaches and why they might matter to technical translators, see also the following Wikipedia links:
Rule-based machine translation
Statistical machine translation
Machine translation
Global Watchtower describes the latest advances of the four main players in this field, PROMT (good quality, integrates with Trados, sadly no English > Italian module yet), SDL (finally offers Trados integration), SYSTRAN (no real integration with CAT tools except with workarounds), and Translution (no integration with CAT tools, as far as I know).
One important point made in the post is the recent shift, supported mainly by Google, from rule-based to statistical as the technology of choice in this sector. Perhaps we will see more blended approaches in the future, as traditional rule-based system providers tweak their programs so that they can benefit from the large parallel corpora of texts that are made available on the internet.
Global Watchtower’s closing paragraph tries to highlight the main benefits offered by desktop-based solutions (programs that you have to buy and run locally on your PC or LAN server) as opposed to web-based approaches, citing integration with the desktop, easier user manipulation of the data and protection of sensitive data.
From the point of view of a technical translator, I cannot see how I can ever be able to boost my productivity by using a web-based product such as Google MT, on which I can have no pre-production control whatsoever. Without a tight control on the terminology that such a system will spit out, its results will be almost useless and require plenty of repetitive, tedious post-editing.
Instead, using a desktop MT system that accepts carefully “hand-crafted” glossaries (perhaps automatically extracted from existing reliable translation memories), a technical translator may be able to obtain considerable advantages both in terms of consistency and speed.
Dear Roberto,
thanks for pointing out this post I didn’ see on Global Watchtower. Regarding the Systran integration with CAT tools, I don’t know if the version 7 can import or export SDL Trados files, but this very last version combines rule-based and statistical machine translation in a very promising way. And once your text is translated by the engine (which applied both techniques), it can be edited in a TM editor. Each segment you decide to approve might be locked (the source will never be translated again) and / or sent in the corpus which is used to train the statistical engine. I attended a presentation they made at Paris-La Défense, and I will soon write a comprehensive post (on my blog, L’Observatoire de la traduction about Systran Translation Server 7.
Best regards,
Guillaume de Brébisson
Site de traduction Anyword
Hi Guillaume,
thank you for sending the information and link to your website. I’m looking forward to hearing about the new features offered by Systran 7.
Guillaume bonjour,
je me permets de vous écrire, car j’ai lu pas mal de choses sur les traducteurs automatiques et je dois faire un choix pour mon entreprise. je vous avoue que je suis perdue.
pourriez-vous m’orienter dans mon choix.
merci par avance,
Rodica Sinic
Hi, i think that i saw you visited my weblog so i came to “return the
favor”.I am attempting to find things to improve my
web site!I suppose its ok to use some of your ideas!!
My site – generateur de code Spotify