Which features are translation memory applications missing today?

Translation memory (TM) applications such as TRADOS, SDLX or Star Transit are products that have become just as familiar and commonplace as word processing applications for translation agencies and translators. There are several TM software applications currently available on the market, but all offer more or less the same functionality.

As such, what can we envisage in terms of improvement? For years, these products have not really evolved in any significant way. Nonetheless, several things could be done.

For example, translators work most of the time from home. Today, there is only one way for several translators to collaborate on the same project: they must all relocate to the same site to share one central translation memory. However, it would be much more convenient if they could stay home and share remotely one central memory to benefit from each other's work. It would also be useful if they did not have to worry about which TM application is used or preferred by any given agency or client. Indeed, the translator should not have to be a computer expert, required to learn and master different interfaces for various tools.

It would be ideal to be able to connect to existing TM software applications to take advantage of the functionality currently lacking, such as:

• accessing and sharing TMs remotely, via the Internet;
• being able to access several memories at the same time;
• being able to define access rights to memories.

Accessing and sharing translation memories

Why is it important to be able to share remote translation memories? For the sake of illustration, let's imagine a hypothetical translation company, Xiatal Inc., which has three subsidiaries: one in New York, one in London and one in San Diego. The New York and San Diego branches use TRADOS, which has been installed on different computers in each office. The London branch, on the other hand, uses the TM system SUPERMEM, a proprietary tool developed in-house by Xiatal Inc. engineers, as well as the SYSTRAN machine translation tool freely available at babelfish.altavista.com.

Each office has approximately one hundred TMs available, as well as a specialist responsible for maintenance, i.e., importing, exporting, merging, exchanging and globalizing the memories. When a project is launched, the specialist selects the appropriate memories and sends them to the different translators participating in the project.

By virtue of their individual TRADOS licenses, the translators proceed to translate the documents allocated to them. There is no sharing of translations. Following translation, they return their memories to the specialist who merges them into a single memory. Thus, there is an enormous amount of time devoted to managing the TMs.

This exceedingly complex process could be substantially improved. Indeed, all the memories could be indexed and categorized by client, type and product within a central repository. To do so, one would need to specify where each memory is physically located, e.g., in London, New York or San Diego, in our example. What is important to remember is that the TM software would be physically stored on the computers in the company's branch offices, and not on the individual PCs of the translators or on any central computer. This arrangement is known in IT as "Distributed Architecture," which has the distinct advantage of retaining system functionality in the event of a breakdown on one or more of the computers within the network.

Likewise, one could specify the type of memory in place (TRADOS TM in New York and San Diego, for instance, and SuperMem TM in London, in addition to the SYSTRAN MT. Why? Simply because the multiplicity of TM systems available today renders the management of these memories much more complex by imposing the need to work with exchange formats (such as TMX) and to tweak and excessively handle the files. Why not, then, try to avoid all this extra management work by constructing a system that could connect to any other system capable of rendering the translation of a sentence?

At the time these memories are indexed, each specialist would continue to manage his or her memories locally, just as before. The innovation here lies in the huge advantage one gains by making the TMs from each branch available for projects in all the other branches. Another major innovation would be to make the memories available online to all translators who would then be able to share their efforts while working from home.

Access to the memories would thus be centralized and controlled from within the projects, while each memory and every translator could be located anywhere throughout the world.

Access to several TMs of different types at the same time for any given project

Furthermore, centralized project management would allow the manager to link as many memories as desired to any given project. For example, one could define one memory as "Project TM" for the translations linked to the project proper, and then a "Client TM" that contains a combination of all translations of all projects for any given client, and finally, an "IT TM" that would contain all specialized translations carried out in the field of IT. These three memories could be read and updated at the same time during the course of any given project. One could even define which TM would carry preference in the case of an identical match arising from two TMs.

A Project Manager in New York could thus create a new project by using, for example, a local TRADOS memory, another TRADOS memory from the Tokyo office, and a third SuperMem memory from London. These memories could be of different types within one and the same project.

Access rights

Another problem often encountered when using the current systems is the risk of having a high-quality TM corrupted while in use for a project. With no quality control mechanism in place, translators are able to introduce new translated segments of poor quality into TMs, or worse, modify translations already validated by a client.

This problem could be avoided with centralized TM management. For example, a Project Manager could decide to assign two translation memories to one project, one in Paris and one in New York. Then, knowing that the New York TM had been validated and should not be modified, he could set it up as read-only to ensure that it would not be manipulated. In that case, the translators would receive input from both memories, but only the Paris TM would be populated by new translations, leaving the New York memory unchanged.

Without going into excessive detail, a very sophisticated system of access rights would make it possible to allocate group rights, local rights, or access rights, for example, only to certain groups of senior translators, (and not junior translators). Senior translators would be authorized to make changes to the TM throughout the course of the project.

Simplifying project setup

As soon as a project has been properly configured, sending memories to the individual translators by email or ftp would no longer be necessary. It would suffice merely to send them the project code, user name and password. This would be a huge timesaver! The translators would automatically have access, via the Internet, to the memories defined for the project, as if those memories were on their own local computer drives.

For example, as soon as a translator would complete a sentence in Brussels, it would be recorded in the online TM and would thus be available almost instantaneously for his colleagues working on the same project in Paris or Moscow.

This only underscores the fact that translators could be located anywhere in the world sharing their translation memories, which likewise, could be located anywhere on the globe.

Again, there are significant gains to be had. Financially, considerably less money would be expended by companies who would benefit by leveraging the number of words emerging from the memories. Likewise, consistency in terminology during the translation would be maintained.

No training costs!

Centralized project management, as well as the possibility of querying different TM systems, would yield a particularly interesting secondary benefit: a standard interface to be used by translators no matter what TM application was linked to the project.

This could be a decisive asset for any translation service bureau or company contracting or outsourcing to many translators. In fact, a company could decide at any moment to change their TM system without needing to change the client interface! For example, our hypothetical company, Xiatal Inc. could decide to migrate all of its TRADOS memories to the SuperMem system (or the reverse). With a system such as we have envisaged, the migration would imply no change whatsoever at the level of the user interface for the translator. None at all! He or she need never know that Xiatal Inc. had changed the system, since the interface would remain exactly the same.

Confidentiality of translations

Remote access to translation memories would significantly improve data security since the memories would remain under the control of the client or service bureau involved. There would be no need to distribute the files by email or ftp, as is currently done everyday by translation agencies and companies.

Instead, communications would be encrypted for dialogue with remote TMs, just as for secure electronic commerce sites, thus guaranteeing confidentiality at each step of processing the data.

Improvement in project management

Moreover, since this could all take place online, a statistical module could also continuously record how many words had been translated for each project and by which translator. This would offer the following advantages:

• the possibility of beginning the review process without waiting for the translator to finish, since the memory would be available on-line;
• the possibility of having the Project Manager effectively track the number of words translated by the translators. Indeed, as soon as a sentence was translated, the statistics would be available through Web-based administration tools and, more importantly, automatically updated

In response to the market's requirements, Telelingua Software has created T-Remote Memory. T-Remote Memory has its own TM SQL engine. The software complements existing TM applications while adding functionality. A unique advantage of the software is its ability to integrate different TM systems so that individual project's and company needs are met. For more information, please visit www.telelingua.com.

Philippe Mercier is Managing Director of Telelingua Software. He is the author of ten software programming books, published in France by Marabout (Hachette group) and can be reached at pmercier@telelingua.com.