I had promised to write about MultiTrans a few editions ago, so I updated my long-expired trial license from earlier this year and reinstalled and tested and . . .  really did not get anywhere. Fortunately, I've found out in the meantime that much of this was my own fault. Still, I ended up asking to talk to them rather than continuing to meddle around unsuccessfully with their product. (By the way, a few years ago I actually used this tool for a couple of projects, but it had simply changed a little too much for me to be successful in trying out all its new features.)

So let's look at MultiTrans in general and the newer features in particular.

MultiTrans is a TEnT, a translation environment tool, that has a slightly different slant (but then, don't they all!).

Rather than relying on a traditional translation memory, it relies on something called bitext (or "corpus"). The difference between a TM and a bitext is essentially that a TM is made up of unrelated sentence pairs without any context, and a bitext consists of complete document pairs that naturally present you with all the context you would like. (A third database technology is by reference materials for which all your old file pairs have to be kept and you can utilize and index to automatically locate matches within these file pairs -- this is the technology that Star Transit uses.) Back to the bitexts. An interesting development in the differentiation between traditional database-based TMs and bitexts is that the lines have sort of blurred. At this point, most TEnTs use a feature called context matching, 101% matching, guaranteed matching, or perfect matching, which is essentially a way to assign a 100% match a higher match likelihood since it is surrounded by the same context. For this the context naturally has to be stored in the TMs, making them strikingly similar to bitexts. On the other hand, MultiTrans has lately also embraced some principles of traditional TM technology, for instance by storing matches of the current project in a temporary TM that will be kept until the project is finished (and the file pairs are at that point entered as bitext). Because of this mixture of technologies, MultiCorpora, MultiTrans' parent, has renamed its technology to "TextBase TM."

Another differentiator in which MultiTrans excelled was its subsegment leveraging. Through an intelligent combination of information for the terminology database and advanced indexing of the bitext, MultiTrans was early on able to locate subsegment matches, a technology that it is still trying to advance (in its present lingo MultiCorpora calls it ALTM -- advanced leveraging translation memory). Of course, with a number of other tools (including Trados and MemoQ) also supporting subsegement leveraging, this feature has encountered some competition.

The last big distinguisher was the alignment feature. Because MultiTrans uses the document-based approach rather than the segment-by-segment approach of the typical TEnT, the speed of the alignment is -- comparatively speaking -- lightning fast and the accuracy is, well, different than in other tools. Since there is no manual element to the alignment process, problems do occur, but these can be fixed once they are located during the translation. Remember, the materials are stored as complete documents, so it's no problem to quickly re-align a text on the fly once an error has been spotted and corrected. That said, WordAlign, one of the new features of MultiTrans, has greatly improved the accuracy of the alignment. This process creates and extracts glossaries during the alignment that it then further uses to assure that the ongoing alignment gets better and better (this is in addition to existing termbases you can select to enhance the alignment). No other TEnT can do anything like this, either in speed or in accuracy, and the only two tools that match MultiTrans' performance are (you, loyal newsletter reader, will already know what's coming) the fellow Canadian alignment tools AlignFactory and AutoAligner.

All right, maybe we should actually say something about the tool itself and not just its features.

MultiTrans is a complex tool with what I perceive to be a less-than-user-friendly workflow, but it's nothing that you can't get the hang of after playing with it a little bit. It works in various translation environments. MS Word is still its preferred third-party environment (alongside PowerPoint and WordPerfect), meaning that you actually sort of perform the translation within these tools. There is also a standalone "XLIFF Editor" which supports a whole range of additional formats (such as HTML, XML, FrameMaker, InDesign, and XLIFF), but it needs to be licensed independently.

When I say "sort of perform the translation within these tools," I mean that unlike the early versions, menus in Word, etc., are still used to connect to the MultiTrans data and processes, but now a little "Translation Agent" sits on top of Word, PowerPoint, WordPerfect, and (from the next release on) XLIFF Editor where you actually type the translation, search for data, and send data to, say, the terminology database. Once a segment is translated it is then entered into the Word or other document in the lower half where you can see everything in its full formatting (the only editor that does not allow for a WYSIWYG is the XLIFF Editor -- here the translation segments are displayed in a table format with only limited formatting on display).

If you need to work in the actual MultiTrans application to align texts ("TextBase Builder"), maintain bitexts or terminology data, use some of the available project management features, etc., you will first of all have to get used to a "different sense of logic." (For instance, if you want to align a large number of files, you will first have to open the completely separate ListBuilder application to match file names, a list which you then save and load into the main application -- yikes!) But you will also quickly find that there are a good number of power features (for instance, you can schedule alignment processing during non-working hours) that point to the most important client focus: large national and multinational organizations. Many of these organizations (and I am talking about institutions like the Canadian government, the African Union, or organizations of the UN) are set up with large in-house contingents of translators and a more limited number of freelancers. While the in-house translators naturally use the translation environment that MultiTrans provides, many of the freelancers don't (according to MultiCorpora's estimate, only 10-30% do). The rest use the bilingual Trados .doc/.rtf format which MultiTrans supports as an interim translation output format (and as a format that can be translated within the tool as well).

Don't misunderstand me: MultiTrans is set up to be used by all sectors of the language industry -- clients, LSPs, and freelancers -- but the latter two categories have just not been the main marketing focus so far. MultiCorpora assured me that this will change, but we shall see.

Other features that are helpful (but again, maybe more for larger organizations) are the integration into content management systems such as eDocs or Documentum, a close integration with the MT engine Systran (to be purchased as a standalone), and the close partnership with the workflow environment Plunet that I've mentioned in previous newsletters.

Overall this is a very, very powerful tool that in some areas is the most sophisticated of all the TEnTs -- or at least can match what its competitors have to offer. I look forward to a new focus on the freelance and LSP markets, which I hope will not only take the shape of marketing but will also include some usability studies and streamlining of some of the processes.