Efficient work on localisation - including strong reduction of the work-load and higher quality translations by translators with little experience - can be achieved by following a clear process for localisation within a well defined workflow.
Process and workflow can be defined and implemented using standard data formats for localisation, accompanied by translation management software capable of managing this process and enforcing workflow.
The use of specific workflows in localisation - in which the different steps that must be followed by a translation work are clear, and in which information is kept on what is happening at each step of the process – lead to:
The goal of storing this information is
A complete definition of the localisation process requires:
The roles that the different players might play in translation are:
A phase is a period of time in which one single type of operation is performed on a set of translatable data. It might be a long phase (such as in translating all the contents) or a very short phase (approving the contents for use). A phase is never interrupted by other phases, as soon as a new phase starts (for example review after translation), the prior phase is terminated, even if later a similar phase (more translation) might be started.
The different roles participate in specific phases:
This leads to a list of phases that might not be exhaustive, but so far covers all the manual phases presently considered in the workflows that we have analysed or know of. The list is:
In addition to these manual phases, there are other processes that can add information to the translation. Some of them can be considered clearly as phases, while other are computer-assisted translation tools which support the work within a phase. These processes include:
The first process directly affects the target field of the translatable files, and should definitely be considered as a phase, as it requires review by the translator or reviewer.
Processes that add information to the file are important to track when they are done in advance of the moment in which they need to be used. As an example, inclusion of TM and Glossary can be done as low priority computer processes that take place whenever the computer has time, but... the data might become obsolete if new TM information becomes available, requiring repetition of the process before the files are handed to the translator. Processes of enrichment of the translatable data should also be considered as phases.
The automatic review of translations - when used for showing data on the quality of the translation to the translator through user interface - is just part of the translation process, operating within the translation phase, and therefore cannot be a phase by itself. If the results are stored in the file itself, in order to pass the information from translator to reviewer or from reviewer to approver, then this could be considered as a phase.
Extraction of translation memory data or terminology from translated files does not really affect the localisation process itself, and should not be considered as phases in the localisation process.
Adding this information about automatic processes, a more complete list of phases would include:
A workflow is the combination (lineal or in parallel) of one or more of these phases. A given phase can be repeated inside a specific workflow (for example, a translation might be reviewed by several reviewers at the same time).
A workflow not only gives the different steps that a translation must follow, in order to be considered complete and ready for the customer, it also needs to define which rights are given to each role.
For example, a translator working alone has in his/her workflow only one translation phase, and no more, but he must have the right to approve the translation, as a non-approved translation cannot be sent to the customer. In a more complex workflow, the translator might only have the right to create and edit the translations, but not of approving them. The difference between an authoritative reviewer and a non-authoritative one is simply having or not the right of editing the translation. A non-authoritative reviewer only has the right to add comments, but might not touch the translation.
What are the rights that might be considered in workflows?
A workflow is simply a sequence of phases that translatable content will follow during its translation or localisation. Each phase is assigned to a player, together with the rights that this player has in the work process.
Workflows might be as simple as one single translation phase by a translator who has the right to approve, or quite complex, requiring several phases of review, some of them mechanical and other performed by human reviewers.
A flexible workflow definition system must allow the administrator to define the phases that the process must follow, then assign rights to each type of player. Allowing choosing between a number of predefined workflows strongly simplifies the work of the administrator.
Examples of workflows might be:
Translator + Non-authoritative reviewer + Editing + Approver. Translator translates. Reviewer makes recommendations Translator edits. Approver approves and sends to customer.
The implementation of the framework described above requires a computer system capable of the following management tasks:
On top of these features, the system should also perform other computer-assisted translation functions, such as glossary and translation memory management, testing, statistics and definition and follow up of goals for teams. The data required to manage these functions includes:
If we define a set of translatable files that need to be translated to a given language as a “project”, the last two blocks of data reference the whole project, while the first block corresponds to each individual file. In this last case, keeping process information in the files themselves allows steps in the translation process that do not necessarily have to pass through the translation management system (such as a translator handing a file directly to a reviewer).
The only standard file format that has the necessary structures to handle process is the XLIFF format. Inside its XML structures it allows for the definition of phases and the assignment of a given message to a specific phase. This permit maintaining historical information about all the steps (phases) that the file has gone through. This information will be used by the translation management system to follow the required workflow for that project. The next section defines how the process data mentioned above can be stored in XLIFF files.