Localization Technology

When you start thinking about translation and localization pro-cesses, you will sooner or later have to deal with the tools and technolo-gies that are being used in this field.

Here are some of the tools you should know about. You don’t have to own the tools yourself, but you should know what they are so that you understand your translation service provider when it is talking about segments, translation units (TUs) and translation memories (TMs).

Today, translators use many tools to help them remember translations of terms, sen-tences or phrases — be it within the same set of documents or between projects. The most important tools here are TM sys-tems, terminology management systems and software localization tools.

Let’s start with one of the smallest items there is in translation: the segment.

Segment: A segment is a translatable unit, usually a sentence, but it could also just be a three-word heading or a single word in a table cell.

Segment pair/TU: This is a segment plus its translation. The pair is stored for later re-use inside a translation tool. If the transla-tor needs to translate the same or a similar segment later on, the translation tool will retrieve the translation of the segment for the translator automatically.

Translation tool: A translation tool is a software program that helps translators remember any segment pair that they have ever made. There are TM systems designed for the translation of documents such as manuals, training materials, cata-logues, websites and so on. On the other hand, there are software localization tools that are used for the translation of the user interface texts of a software product.

Basically, TM tools and software local-ization tools do the same thing: help the translator go through the translatable text segment by segment. It is just that the developers of the translation tools have traditionally focused on dealing with docu-mentation files or software files. With the next generation of translation tools, this difference will disappear more and more.

Software localization tool: As the name implies, these tools deal with files from software projects, such as EXE, DLL, resx and so on. The current family of localization tools can also be used to translate HTML and XML files. In addition to translation, the software localization tools also offer ways to localize or adjust the layout of the software to be translated to better suit a different audience. For example, buttons can be resized so that the translated text fits onto it. Unfortunately, when you start translating with a translation tool of any kind, the system is empty and has to be filled by the translator. Before any re-use can occur, the translator will have to input lots and lots of translations for text segments. If, by any chance, there already exists a translation for a document or software piece, then these translations can be imported into the translation system via an alignment, even if they are not created with a translation tool.

Alignment: Before starting to fill an empty translation tool with new translations, already translated texts can be recycled. An alignment tool will read in the source text and the translated text from almost any file format (except PDF) and show the segments from the documents in parallel.

The system will try to connect the segments from the source language text to those of the target language text, but, as most align-ment tools do not “understand” language, they will only be able to take a guess on which segments belong together. Ideally, a translator or at least a person who knows both languages will have to look through the aligned segment pairs and determine whether they really belong together. After this check, the segment pairs will be sent to the translation tool for further re-use.

The translation tool will now be able to compare each new segment for translation with all segments in the translation tools repository and will retrieve the transla-tion for any identical or similar segment as suggestion for the translator. A transla-tion tool is not able to translate by itself, like a machine translation (MT) tool, but can only re-use what has been saved to the system.

MT: MT systems can analyze the source language text and translate it, with the help of dictionaries, into another language. This works best when the source text is written in so-called controlled language, which lim-its the terms that can be used and simpli-fies the structure of the sentences so that the machine can more easily deal with the text. MT systems are designed for a specific language pair and need large dictionar-ies for the subject matter area they are to translate. Often, either pre-editing or post-editing of the text is necessary.

Terminology: Terminology is a very im portant concern for the translator, not only for MT systems, but also for any kind of translation. A term may consist of one word or a combination of words. Terminol-ogy lists and terminology databases that are connected to translation systems help the translators to achieve a correct trans-lation. Terminology, however, is more than just words; it contains information such as which product the term belongs to, who entered or approved the term, the source for the translation, definitions, and con-text examples. In order to build up a termi-nology repository, first the terms need to be collected or extracted out of the text.

Terminology extraction: While manu-ally collecting terms is the best way to get what you want — company specific terms, subject matter specific terms, new terms and so on — it is also the most tedious. Terminology extraction tools can help to run through a large number of documents to either extract all or the most frequent terms in one language or even from two languages, if the source material is bilin-gual. The lists of terms that are created by such an extraction tool are the term candi-dates — possible entries for a terminology database. Still, the terms you want need to be selected from those lists by a human.

Terminology database: Each translation tool has a way to check text that needs to be translated for the terms, be it with inte-grated terminology databases or term lists that are loaded into the tool. Terminology management is an ongoing process, involv-ing adding new terms for new products, changing terms, adding forbidden terms, or updating additional information such as specific uses for a certain product or a cer-tain customer. It involves the management of definitions, context examples, notes and graphics.

Terminology check: While terminology databases and lists are provided for trans-lation, it sometimes might make sense to check whether the terms from the list have been used correctly and consistently, or whether or not forbidden terms have been used. If the terminology database is set up in a way that shows approved and forbid-den terms, the translation can be checked for the occurrence of either. These checks are integrated either into the translation tool itself (or its terminology component) or can be achieved through dedicated checking tools. Term checks are also avail-able for the source language documenta-tion as there are term checking tools that can connect to the editing environment.

In addition to all these translation-spe-cific tools, there are many more that might be needed:

Ÿ           Conversion tools and text extraction tools that convert/extract text from the original file format to one that can be used during translation.

Ÿ           Word count tools to determine the amount of translatable text and, if they are part of a translation tool, also to determine how much re-use (for example, how many matches) there will be from the TUs in the translation tool.

Ÿ           Quality assurance tools, either belong-ing to the translation tool or as standalone tools to check for missing punctuation, integrity of numbers, spelling, grammar and so on.

Ÿ           Project management tools to create project packages that contain the trans-latable files, reference material, term lists and TUs for re-use in a translation tool. They keep track of the files and their sta-tus (translated, proofread and so on),  the schedule and who translates what.

Workflow management tools are used to automate steps such as converting a file to translatable format, counting the words and checking a translation tool for re-usable TUs (matches). With online workflow man-agement systems for example, translators can be notified as soon as a file has been assigned to them automatically.

本文来源：Guide to Going Global 2008，Multilingual Computing