Any freelancer contemplating the purchase or upgrade of translation memory (TM) software must consider a package that addresses his or her own needs while interfacing effectively with the TM databases employed by large language service providers and corporations. With the exception of the unabashedly modest Wordfast, well-established TM software retails for hundreds of dollars. But is there anything really stopping a TM package from being powerful, widely compatible and yet moderately priced? New TM offering Heartsome seems bent on answering with a resounding no.

This Singapore-based contender first appeared in December 2002, with its current versions and pricing introduced in early 2005. In a real industry first, Heartsome works across platforms (Windows, Macintosh, Linux) in all languages; supports major third-party databases (DB2, MS SQL, Oracle, PostgreSQL, MySQL) plus all key industry standards — XLIFF for formatting, TMX and TBX for memory and terminology database exchange; and is currently working on SRX for segmentation and the emerging word-count proto-standard GMX.

In pursuing its vision for the future of TM, Heartsome has chosen to lay solid design foundations that should provide it with continuing performance answers in a rapidly evolving market, in an approach that mixes commercial sense with a whiff of globalization subversion. The edifice growing on these foundations is still incomplete in various ways, but we are dealing with a work in progress whose gaps are even now being plugged. Indeed, the product is growing apace, thanks in no small part to the product’s interactive website where users can post comments and wish lists to be addressed by the designers.

So, while the program’s interface may not be as intuitive (yet) as that of the more established ones, its solid operating foundations allow translators to work directly on XLIFF and TMX formats rather than on proprietary formats that then have to be converted into and out of XLIFF or TMX. This is a real departure from the norm and the kind of broad-based anticipation of future needs that lends weight to Heartsome’s claim of providing “new generation language solutions.”

The Heartsome Translation Suite

Having established that Heartsome is undeniably standard — and developer — friendly, the next question is how useful it is. For this review we trialed the Heartsome Translation Suite, with all testing done on Windows only.

The suite is composed of three programs: an XLIFF Editor in which source files are converted to XLIFF format and translated; a TMX Editor that deals with TMX files; and a Dictionary Editor that deals with TBX files.

XLIFF Translation Editor. A standard that not many translators would have heard of a few months back, XLIFF (XML Localization Interchange File Format) is fast gaining relevance, having already been adopted by the localization tools CATALYST, PASSOLO and RC-WinTrans — and, more recently, by SDLX. By allowing for the separation of text and formatting and then their post-translation merging in a way that is tool-independent, XLIFF gives large corporations more control over the localization process in terms of both the number of files to be sent for outsourced work and the management of the linguistic assets stored in databases. Accordingly, XLIFF will likely get support from major users and in the current fluid climate can be expected to quickly supersede the previous TRADOS “uncleaned file” de facto standard. After all, TRADOS’ new owner already supports XLIFF.

Heartsome comes with filters that allow for conversion into and out of XLIFF in text, HTML, XML, InDesign, MIF and other formats. The big and rather radical exception in file compatibility is Microsoft Office documents. Word needs to be converted to RTF first, while Excel and PowerPoint require OpenOffice to be installed. This rather cheeky tweaking of the Microsoft giant’s nose is one example of that subversive element we mentioned earlier, and it puts Heartsome on its own as the only commercial TM to dare to snub the world’s main document creation and editing suite.

The XLIFF Editor layout is interesting. Translators are used to two kinds of TM environments: the TRADOS/Wordfast model in which the target segment appears below the source segment and the Déjà Vu/SDLX model in which source and target appear horizontally aligned on a vertical table. Heartsome opts for a new approach where the screen is divided into four windows, with the left side reserved for the source segment and its translation and the right for database matches in source and target languages. Icons and hotkeys allow for the usual navigation between segments and matches.

From this editor, single files or whole projects are converted into and out of XLIFF format, and databases are created, removed and imported/exported from/into TMX or TBX formats. The XLIFF Editor does not, however, allow the user to open and merge or split databases. These operations require the use of the package’s TMX Editor.

TMX Editor. The TMX Editor fulfills functions that in other products correspond to alignment and database management tools. The alignment tool allows the creation of TM from legacy translations with the output formatted directly as a TMX and thus compatible with any of the established TM products that are at least Level 1.1 compliant.

All the expected features at row level are present: edit, split, merge, delete and spellcheck (this requires separate installation of open-source Aspell dictionaries). These same features are also used for database management.

While the TMX Editor also allows for splitting and merging databases, it lacks a basic facility of most other TMs in that database contents are not viewable inside the program itself. So, for instance, you cannot check why a particular sentence you feel must be there does not appear when you click the corresponding icon. The only way to view a database’s contents is either by using the tools that come with it or by exporting it as a TMX or TBX file — which must then be re-imported if any changes are made.

TM applications work with two kinds of databases, one for sentences and another for terminology. Heartsome calls its sentence database TM and its term (and phrase) database Quick TM. Under the Heartsome philosophy, the difference is not so much one of term and sentence databases as of TBX and TMX files. Hard terminology data is kept in TBX format that can be mixed with soft phraseology information in the Quick TM database. This means that translators happy to use glossaries with just minimal information will only need the TMX Editor. If, however, the more sophisticated TBX format is required, then the third component of the suite comes into play: the Dictionary Editor.

Dictionary Editor. This should have been called the TBX Editor: it allows you to create and edit TBX files that can then be imported into the Quick TM database. Again, the only way to view the database contents is to export it as a TBX file. Data can be entered into a TBX file through two customizable templates: simple.xcs and full TBX.

Term extraction is another interesting feature that not all TM applications provide, whether integrated or as an add-on, and Heartsome does. It does not yet allow for batch processing of the files from which to extract the terminology. Each must be processed individually. This feature also includes a ready-compiled and easily customizable list of stop words.

Product evaluation

For this review we downloaded Translation Suite 1.0-9b on Windows XP Home Edition. Java Runtime was already in the computer, and installing Heartsome was as easy as double clicking on the downloaded icon. Documentation was succinct and clear, although not very comprehensive. We browsed through the sample data provided and then tried samples of our own (Word and HTML files). There was a learning curve, but, on the whole, it was comparable to that required by any other TM program.

Importing the files. The first step — and the first hurdle — is converting a single file or the several files that may comprise a project into the XLIFF format. In order to convert a file into XLIFF, one needs to know the File Type (whether RTF, HTML or other approved format) and also the Code Page, which might be more difficult. In most instances, one hopes, the program automatically detects what this Code Page is (whether UTF-8 or Windows 1252 or any other of the over 100 possibilities). In the few cases where it does not and particularly for HTML, using Notepad to check the file header normally reveals the Code Page identity. If all this plus simple trial end error fails, then a call to the client may be necessary. This idiosyncrasy of Heartsome can be annoying, but you may find compensation in the form of entry into the XML environment — precisely where the future of multilingual content is most likely to lie.

Save for these gripes, everything else is straightforward. At this stage you can select languages and databases. Segmentation can be based on the paragraph instead of the sentence, if so desired, which may be convenient for some texts such as literary or advertising material. There is also the option of opening the XLIFF file in the editor once it has been converted.

What users expect from an editing screen is a pleasant, familiar interface that gives maximum context with minimum clutter and allows the user to move ahead with a minimum number of clicks. With Heartsome, the original four-window screen’s saving on clutter is offset by a loss of context, with only the current translation unit on display. But a Context View function allows the user to gain access to the four segments around the working one, giving roughly the same context to be found in competitors’ vertical or grid editors. Then there is a much more comprehensive Preview Translation in HTML — similar to the Déjà Vu external view but without the ability to edit on it. However, it’s still not the WYSIWYG view offered by the Source/Target Preview of the Tag Editor in TRADOS, for example.

The editor allows the translator to do the usual TM editor operations: go to next segment, approve translation, copy source to target, add notes and so on, but often the processes involve more clicking here. Inline formatting in the XLIFF file is represented by easily identifiable green number tags.

While even the no-frills Wordfast allows you to simultaneously use two memory databases and three glossaries, the XLIFF Editor only caters for two databases: TM and Quick TM. Although other databases can be consulted through the search icons, the issue here is that the application is not actively looking at them automatically for matches. Also, words need to be copied and pasted into the search window, a process again involving too much clicking.

When more than one match is retrieved from the databases, the Matching Translation window only offers them one at a time, starting with the closest, with the others accessed in turn by clicking the relevant icon or via hotkey. In a way this is fine — it leaves the space uncluttered — but when matches from both the TM and the Quick TM databases appear, many users would surely prefer to have at least the first from each. Another clearly desirable feature would be a shortcut to enable a term leveraged from Quick TM to be entered at the cursor point, instead of having to import it along with the rest of the source segment or actually type it.

Database Management. In Heartsome’s world, the alignment tool is treated as an integral part of database management. In a way this alignment feature is even more powerful than that of TRADOS or SDLX, in the sense that it allows for genuine batch processing without the need for handling the checking file by file. It does not seem to use, however, anything more than natural order (the position of untranslatable strings such as numbers, for instance) for segmentation accuracy, unlike the more sophisticated approaches of LogiTerm and, to a lesser extent, WinAlign.

The usual merge/split, remove/add functions are present and are as intuitive as in any of the other translation tools available, if not more so (note that this merge/split feature does not exist in the XLIFF editor though). Output goes straight into TMX, with no other intermediate, proprietary code.

As has already been remarked, one major drawback is that translation databases are black boxes that remain impenetrable unless exported as TMX or TBX files and then opened in the corresponding TMX or Dictionary Editor. This may be particularly important for beginner users when, as usual, something unexpected happens and one has to troubleshoot it. This is obviously a minus, but on the plus side, in a networked version the user can choose the database format rather than being obliged to settle for a proprietary one. The desktop version uses IBM Cloudscape, also known as Apache Derby.

Quick TM. As mentioned, Heartsome stores glossaries and terminology under the name of Quick TM, and this can include files with a TMX extension or a TBX extension. Indeed, both types of files can be merged in it. The way to enter legacy glossaries is by saving them as CVS files, then converting them to TMX.

Only the Dictionary Editor is able to open TBX files, but conversion from TMX to TBX or vice versa is a one-click affair using the Export As icon. This means that the freelance translator with no real interest in lexicography can skip buying the TBX editor altogether. Such detail is not often needed, and the time to enter it is rarely available. Meanwhile, detailed terminological data received in TBX format can always be entered into the Quick TM.

Entering terminology and phraseology on the fly is also possible, but the procedure is cumbersome and lacks the simplicity of highlighting the words and clicking once (SDLX or Déjà Vu) or twice (Wordfast). In Heartsome the terms have to be typed or copied and pasted in the opening window. Any other data must be entered after the file has been exported out of the database.

Dealing with projects. Batch processing of files is possible as long as they belong to the same File Type. The user can then merge (and split) the resulting project files and access the Repetition Analysis feature, which gives information on internal repetition and available database matches, then on segments that have already been translated. This feature does not work with individual files. The Accept All 100% Matches command is the closest Heartsome gets to a pretranslation feature.

As for cross-compatibility, small TMX databases produced by Heartsome were imported into other TM packages without problems, and it successfully imported TMX files from them. Heartsome also correctly processed TRADOS files and vice versa.

Conclusions

A quick look at the Releases History section at www.heartsome.net illustrates just how much the program has advanced in 2005 alone. The pace is such that many of the issues discussed here may have been resolved by the time this review goes to print, with the launch of Heartsome’s Release 6. Promised improvements for that release or shortly thereafter include XLIFF Editor to implement split/merge segments; search in TM populated instead of requiring manual entry; and improved interaction between background databases and the databases in use. For the TMX Editor, direct database editing without the need to export first. For the Dictionary Editor, a much-needed Getting Started guide and a bilingual term extraction from TMX files. Technical support is attentive and rapid, and users are encouraged to become actively involved in the product’s development through an internal mailing list and a wiki.

Heartsome’s pricing is another surprise. At US$88 for the XLIFF Editor plus US$68 for the TMX Editor (as noted, many freelancers could do without the Dictionary Editor), it is cheaper than Wordfast while offering all the features of freelance versions of major TM brands. The price issue doesn’t end there, however. By running Linux and Open Office, one can save on all the underlying costs for the latest versions of Windows (or Macintosh) and Microsoft Office while still being able to work around the ubiquitous Office framework. This could prove a real boon for translators in developing economies, for example, or students who may be cash poor but “time rich.”

It’s worth remembering that the translation profession has a goodly number of older practitioners for whom computers have come along relatively late in life. However, there is now a generation who has practically grown up with them, and it’s to this newer, IT-savvy crowd that Heartsome will presumably have broader appeal. Such users would appreciate the software’s development philosophy (including, as we have observed, a subversive little tilt at Microsoft) and be better equipped to exploit its hands-on, multiplatform, multi-database nature. For others not so versed in computing and for whom acquiring TM proficiency has been something of a via crucis (witness the agonized pleas on some of the TM user group lists), Heartsome with its embarrassment of choice could prove a little complicated and, well, geeky.

As for corporate users, the future looks to be with XML, and major translation agencies and clients may like the greater control that Heartsome offers over the whole translation cycle through XLIFF and TMX. The fiddly aspects we have seen might temper their enthusiasm, and it will be interesting to see whether any of the existing major brands sees Heartsome as a potential threat and incorporates some of its strengths without its current weaknesses.

In summary, Heartsome is a headstrong little package that clearly means to go its own way. It is built on a solid basis with an eye to the future, and its rough edges are being smoothed by active developers, with users also having a say in the process. Not everyone will want to invest the time on it, but time is about the only real necessary investment. For translators on a budget, the combination of Linux, Open Office and Heartsome neatly sidesteps all the big guys and provides TM power for a pittance. If you have a computer, now you can play too — commerce and subversion indeed. M