Public Services and Procurement Canada
Symbol of the Government of Canada

Institutional Links

 

Important notice

This version of Favourite Articles has been archived and won't be updated before it is permanently deleted.

Please consult the revamped version of Favourite Articles for the most up-to-date content, and don't forget to update your bookmarks!

Search Canada.ca

Assessing translation memory functionalities

Julie L. Gariépy
(Language Update, Volume 9, Number 2, 2012, page 26)

Translation is a dynamic discipline. Recent technological innovations are changing the way people translate. New tools, such as translation memories, are designed to speed up the work process and ensure consistency in individual and team translation projects. One thing is certain: translation memory systems are here for good.

We should not forget, however, that translation memories are not all the same. There are as many memories as translation tasks. Therefore, before assessing a particular tool, it’s important to consider whom the assessment is for and why it is being conducted. To that end, a list of requirements for a specific task and user or group of users can be drawn up.

This article provides information for assessing the functionalities or features of translation memories. I’ve included some parameters to consider and suggested ways of testing them. Since I’m not assessing a particular application, no specific user or group of users is targeted.

Compatibility and consistency

Translators are called upon to translate various types of documents (presentations, reports, certificates, websites, etc.), which are generally prepared using specialized software. They therefore receive files in many different formats from their clients. A good translation memory should be able to process all the file formats in which the translator using it works. It should be able to import and process files in Word, WordPerfect, HTML, PDF, PowerPoint, etc., and allow translations to be produced in compatible formats so that the translator doesn’t have to convert them.

There are also several operating systems on the market. Before buying a translation memory, translators should ensure that the application works with the system installed on their computer. Most memories are currently designed for Windows, but in the future, they will no doubt run directly on a Web browser, meaning they will not be installed on users’ computers, but will be accessible on the Web.

Updates are another important aspect to consider. It’s important to find out whether they are free and whether they are automatically downloaded. If every update has to be purchased, then this should be taken into account in the cost of the application. These days, some applications can process exchange formats (which are not associated with any supplier in particular): TMX (Translation Memory eXchange), TBX (Term Base eXchange) and SRX (Segmentation Rules eXchange). A memory that processes and exports these types of files will save the user a lot of conversion work.

To assess an application’s compatibility, I suggest drawing up a list of user requirements and then checking the product website for the product’s technical information, a description of the necessary configuration and the policies regarding updates. I also suggest testing the memory by importing, exporting and running files in various formats to ensure that it can process them effectively and that their layout does not change.

Access

Translators often work on teams—at a company, for a government organization, with a group of freelancers, etc. The person responsible for buying the application should make sure it can be used by more than one person at a time. If it can, by how many people? Does a licence have to be purchased for each person? If so, is there a discount for translation firms?

Also, will the application give users shared access to the terminology database or individual access to a restricted database? If a team is working on a group project, it would be better off sharing the database to ensure consistency. However, if a translation firm has a designated terminologist, does the application have the option of letting all the translators consult the database, but allow only the terminologist to make changes to it? An application that allows for distinct types of access for each user (e.g. new translators, experts, revisers, terminologists, managers) would definitely be more useful in a translation firm or a government organization than for a freelancer. On the other hand, a freelancer who often works with others could clearly benefit from a memory exchange feature or shared access to a memory for a particular project.

To assess the access features, I suggest consulting the product website or contacting the vendor. It would also be a good idea to visit blogs where problems encountered while networking or sharing memories are discussed. A good application will process the commands of multiple users without its speed or accuracy being affected.

Languages

Some applications process roman characters only. Translators working in Russian, Arabic or Asian languages should ensure that the memory they are using processes these languages. What’s more, punctuation is not the same from one language to the next. An application could therefore make alignment mistakes if it’s not designed to process a certain language.

Language also affects terminology modules. Does the application include an exclusion list for the languages in which the translator works? If so, the terms proposed by the terminology extractor will be more relevant. Can multilingual terminology records be created in the terminology database? And if so, is there an option to consult only certain language pairs for a given project without creating a database for each language pair?

The same questions apply to corpora. Can multilingual corpora be created or imported? And if so, is there an option to consult only certain language pairs for a given project? In the case of bilingual corpora, can the source and target languages be switched?

In addition, it’s often easier for translators to work in their dominant language. An ideal memory will offer an interface in the translator’s language.

Usually, basic information on languages can be easily found on the product website. However, the person purchasing the application could consult online discussion groups that talk about the application’s accuracy when it processes the translator’s languages, since some applications work better with one language pair than others. If a trial version is being offered, the person purchasing the application could experiment with the language pairs of interest.

Size of and changes to corpora

Unlike freelancers, a translation firm will need a lot of storage space for all the documents it translates. If it has many clients, it would be best to create one corpus per client to ensure syntactical and terminological consistency. It’s therefore important to think about requirements before buying a translation tool.

Translators will definitely want to delete obsolete documents from their corpus and add the new ones they produce. When they are working in a group on a new project, does the application let the team members combine their translation memories? And if mistakes are found in a document, does the application allow them to be corrected?

An essential aspect of translation memories that is often forgotten when they are assessed is how they record and allow users to view their administrative data, such as the creation date of a document or terminology record, or a document’s author or translator. In a rapidly changing field, this data is essential for ensuring that the most recent documents are used. The data can also be used by a team to identify the translator who produced a particular document or terminology record.

Information on corpus size should be on the product website. To assess the user‑friendliness of the features for adding and deleting files, the correction features and the accuracy of the administrative data, I suggest testing the application.

In conclusion, assessing the functionalities of a translation memory can be complicated. Before analyzing a memory, users should draw up a list of requirements and then check whether the tool has all the necessary features. The aspects discussed in this article will need to be considered, as well as possibly the different ways of viewing corpus documents (context displays, KWIC format, etc.) and the integration of modules (terminology databases, extractors, word processors, etc.), depending on personal requirements. As is the case with translation solutions, there is not just one single good translation memory. Assessments of applications should look more at which tools are appropriate for particular work contexts.

Sources

  • Asanka Wasala, Ruwan and Ruvan Weerasinghe. “Initial Survey on the Availability of Translation Memory Tools,” Pan Localization, 2007, http://www.panl10n.net/english/Outputs%20Phase%202/CCs/Srilanka/Papers/2007/0702/Research_Report_on_TMs.pdf.
  • Bowker, Lynne. “Translation-Memory Systems,” Computer-Aided Translation Technology: A Practical Introduction, University of Ottawa Press, 2002, pp. 92–127.
  • EAGLES Evaluation Working Group. “Design and Function of Translation Memory,” Multilingual Information Processing Department, ISSCO, University of Geneva, 1995, http://www.issco.unige.ch/en/research/projects/ewg95/node152.html.
  • McBride, Cheryl. “Translation Memory Systems: An Analysis of Translators’ Attitudes and Opinions,” University of Ottawa, 2009, http://vista4.uottawa.ca/webct/cobaltMainFrame.dowebct?appforward=/webct/startFrameSet.dowebct%3Fforward=organizer_generalFromCourseChannelList%26lcid=97914550001.
  • Zerfass, Angelika. “Evaluating Translation Memory Systems,” 2002, http://www.mt-archive.info/LREC-2002-Zerfass.pdf.