Okay, here is the promised article on the second TEnT that was developed with a developer's mind -- and I am sure glad I held off on sending it out last week because it turned out that I really had seen only a small part of it.

So, let's start at the beginning. I didn't have to think much to come up with a title for this article, because the developer of MT2007 -- Andrey Uzbekov, aka Andrew Manson -- must be a very lonely man. Otherwise, I wouldn't understand why he put scantily clad women on the welcome screens of his software. (And if you think the woman in the main application is scandalous, check out the one in the alignment tool!) Seriously, I dithered back and forth on whether to write about this tool since it seemed just a little too much on the side of bad taste. Nonetheless, here it is.

It appears that the name -- MT2007 -- is an abbreviation of a reverse translation of Память переводов -- Memory Translation -- rather than Machine Translation as most of us would suspect. It is a Russian tool, and it is shockingly robust and feature-rich for being donation-ware. Aside from its rather appalling welcome screens, it's a very attractive program that provides a clear and easy-to-use interface . . . that is, if you read Russian or if you manage to switch to the (semi-)English interface (you'll need to run the LanguageSelector.exe in the bin folder for that). It comes with a very large amount of Russian-to-English glossary/translation memory data from the XDXF project (which makes it a very slow - and, in my case, often aborted -- 160-MB download, but I'm sure it's a very welcome resource for Russian translators). These glossaries, or I should probably say "dictionaries," are open-source material -- if you are not familiar with that project, have a look at the XDXF page for your language combination and you might just like what you find -- as are other components of this program, including the English WordNet dictionaries (also very much worth a look!) and most of the programming components. After you have downloaded the tool it does not require any installation; instead, just unzip it. 

(In case you are running a language version of an operating system that does not use commas as the decimal separator, you might not be able to actually use the program right away, since the font in the Translation Editor where the actual translation takes part will be obscenely large and unusable. This is because the font size is coded as a comma-separated value. To fix that, open MT2007.exe.config in the bin folder with a text editor and change the font size in all lines that contain "editor_font_size." And if you think that's frustrating, read the first two sentences of the manual: "As for now, MT2007 can be regarded as a perennial beta version. Which means that errors are guaranteed.")

While the interface of the tool is really simple, it might take some time to get used to it because the concepts (and terminology) are a bit different than what you might be used to. For instance, there is no real distinction between translation memories and terminology databases in this tool; instead, there is a feature that allows you to highlight subsegments, translate them, and send them to the "repository." The repository seems to be a collective term for linguistic materials including TM data -- which, just as in the upcoming version of Trados, uses SQLite as its underlying database system. Another core concept -- and here the developer's mind becomes only too apparent -- is the use of regular expressions. This is helpful if you want to build your own fuzzy search algorithms for languages with a high degree of flexion. That particular feature wouldn't be my cup of tea, but for those who are good with regex, more power to you.

The way you import TMX (translation memory exchange) data is exactly the same way you would import any other file you want to translate -- only with TMX you will send it to the repository after the import. Once the data is in there, automatic searches are relatively fast and there is a host of keyboard shortcuts (that are easily viewable by way of the little help icon in the Translation Editor) to switch between windows and enter data from the matches that are found. Like Déjà Vu, Heartsome, and Swordfish, MT2007 uses example-based machine translation (although I am still not convinced that this is the right term for what it actually does) by combining matches and "correcting" fuzzy matches through deleting certain parts of target segments and replacing others -- if the necessary material is available in the repositories or the other associated language resources.

Like many other tools at this point, this tool also just integrates true machine translation. PROMT, the Russian machine translation engine, is supported if it is installed along with -- surprise, surprise -- Google Translate.

The translation file formats it supports are MS Office 2007, OpenOffice, and MS Word 2003 saved as XML. Once the files are imported they are displayed in a tabular easy-to-work-with format.

To me it seems that this is a tool that would definitely warrant a second look by English<>Russian translators with a knack for regular expressions (and semi-naked girls), but it certainly has the potential to be more than just that. The author of the documentation for this tool, Konstantin Lakshin, has also written an interesting review of it in the publication of the Slavic division of the ATA that you should consult if you would like to dig deeper.

Overall I was impressed with the tool. I liked the concept -- just take a bunch of freely available components and make them into something new and useable. And while the tool might be primarily for one language combination, that does not have to be a bad thing. Too often we narrow our opportunities by trying to do something that fits everyone. And I like the very laid-back way of "marketing" the tool. Like its self-depreciating beginning, the manual ends by asking: