Applications and modules

The basic architecture and processes of translation environment tools are described by (Lagoudaki, 2008, 38-47). Although the boundaries between TM systems and MT sys- tems are becoming blurred (see 1.4.2), it is still important to make a distinction between them. Basically, TM systems cannot produce target-language text; they can only retrieve previous translations and suggest them. Translation retrieval is dened as the process of retrieving a set of translation records from the TM which are calculated to be of potential use in translating the input [...], (Baldwin, 2010, 196). This retrieval is exible because it works not only for verbatim repetitions of segments, but also for similar segments. The similarity value is calculated using various means, see 1.3.6.2 for a discussion and references for further reading. A TM system does not suggest any translation for a segment for which no similar enough entry can be found in the TM. A pure TM system does not assemble translations and is therefore much more similar to information retrieval systems than to MT systems, see (Reinke, 2004, 58-60). On the other hand, an MT system is expected to produce a translation in any case. A TM system has a dierent purpose: it is expected to help the user by supporting translation in an interactive fashion with target language text suggestions that have typically been written or revised by a human translator.

A TM system is usually part of a larger software suite, the translation environment tool. This suite includes additional applications, see (Lagoudaki, 2009, 28). It is not necessary to devote much attention to this already well-described topic. However, suce to say that any translation environment tool needs at least two components other than the TM system: le format lters and an editor. For most le formats (a major exception being MS Word), the translation cannot be done in the original application, e.g. Adobe InDesign, while using the TM system. Instead, the le has to be converted into another format that can be processed in the editor. For this conversion, le format lters also known as format converters are necessary. Their availability can be an important argument for purchasing one product instead of another. However, poor lter quality can be a major shortcoming, in particular (but not only) for newly developed translation environment tools, see (Geldbach, 2010c, 52 and 54) and (Geldbach, 2010b, 54). In addition, le format lters have to cope with the dierent versions of the same application (e.g. Adobe InDesign CS2, CS3, CS4, CS5).

In order to translate the les, an editor is necessary: this can be proprietary or ex- ternal. Proprietary editors are part of the translation environment tool and do not exist as independent applications. In the case of external editors, a third-party application is adapted for translation purposes. The advantage of the external editor is that a familiar

editing interface can be presented to the user. This is why widespread word-processing applications are used (nearly always MS Word, see Keller (2011)). The adoption of an external editor, the primary task of which is not translation, see (Chama, 2009, 37), also poses some diculties, among others, the reliance on a third-party tool that makes it dif- cult to do your own independent planning for your tool, Zetzsche (2007a). Consequently, proprietary editors have been lately favored, see also 1.4.3.

It is of essential importance to distinguish between the translation memory and the terminology database. A translation memory is, as we have already seen, a collection of aligned texts (as a whole or split down to segments) that is usually updated continuously during translation and automatically queried with each new segment to be translated. This is not the case with a terminology database; it is a separate resource that is not automatically updated like the translation memory. Moreover, terminology database entries are usually shorter than complete sentences, yet their structure can be much more complex than a translation memory entry.

It is obvious that integrating TM systems and terminology systems yields benets, and the two components are found in all current translation environment tools. During trans- lation, the user accesses both at the same time. Still, they are populated and maintained in dierent ways. In addition, there is no automatic synchronization: the translation pro- posed by the terminology database is not per se automatically used in the translation memory. The terminology system can also be used independently, outside a translation environment tool for (even just monolingual) terminology management.

A TM system enables translation memories to be created, but does not include them. They can be either populated during translation, obtained from customers or language service providers for specic jobs, or purchased. The rst method requires time and at the beginning will not yield many results. The second and the third have the disadvantage that the content is not self-generated and its quality might not meet expectations. Additionally, the third method involves also costs and is not completely free of copyright issues, see 1.4.7. A reliable way to build up resources is to reuse one's own past translations completed without a TM system. For this purpose, an alignment tool is included in translation environment tools. The source and target le are processed and a correspondence between them (usually at segment level) is established. The results of automatic alignments should be checked as errors are possible. This work can therefore be quite time-consuming. Further limitations apply e.g. to the le formats that can be processed. However, this step can prove extremely benecial if future translations are comprised of updates, revisions or are otherwise similar to the aligned material.6

The management of complex translation projects involving multiple les, target lan- guages, translators, deadlines and resources is an everyday routine at language service providers (LSPs) and language departments of large companies. Given that translation projects entail some specic aspects, full-edged translation management system have been developed, Plunet BusinessManager being currently the best-known, see Sikes (2010)

6_{Translation memories can be also populated by aligning texts from the web, but this is a special type}

and Panzer et al. (2010). Many translation environment tools include a proprietary trans- lation management system, with dierent levels of complexity depending on the target au- dience of the application version: freelance translators, LSPs or corporations. The features of these translation management systems range from relatively simple tools for freelance translators to full-edged workow automation solutions intended for large corporations.

When it comes to the translation of applications and texts meant for the digital world (online helps, graphical user interfaces, websites), the term localization is often used. Den- ing localization is dicult, see (Dunne, 2006, 1), but a working denition is:

The process by which digital content and products developed in one locale [...] are adapted for sale and use in another locale. Localization involves: (a) translation of textual content into the language and textual conventions of the target locale; and (b) adaptation of non-textual content [...] as well as input, output and delivery mechanisms to take into account the cultural, technical and regulatory requirements of that locale. (Dunne, 2006, 4)

For more information on the term locale, see 2.1.1.1.7 _{Consequently, the term localization}

tool has been introduced, see e.g. Esselink (2000), in order to refer to applications that are specically designed for the translation of these resources. SDL Passolo and Alchemy Catalyst are among the most popular, see (Lagoudaki, 2006, 18), but others are marketed too, see Seewald-Heeg (2009). Their basic functionality is no dierent to that of translation environment tools. However, their distinguishing features (see also (Herrmann, 2011, 22) for an overview) aect the le formats processed because they include software source formats that are usually not supported by translation environment tools. Additionally, advanced preview functions are available and window/frame resizing can be carried out during the translation in contrast with most translation environment tools, in which the layout can be usually adapted only after back-conversion to the source format. The quality assurance includes special checks (e.g. accelerators) that are typical of software applications. For a comprehensive description of the issues relevant to the localization process, see Esselink (2000). In fact, translation environment tools now support source formats that were typical of localization tools, e.g. SDL Trados Studio supports Java resources (.properties les). However, despite some convergence and the fact that basic versions of localization tools are sometimes included in translation environment tools, full- edged localization tools are still separate applications.

Concordancers allow translators to search through [...] parallel corpora, (Bowker and Barlow, 2004, 53), and output the searched string in its context as well as the cor- responding passage in the target language. In this respect, they are not dierent from TM systems. However, the search is started manually. Consequently, concordancers serve more as a reference for translators seeking a suitable translation, usually for subsegments. A basic concordance function is implemented in existing TM systems; more advanced functionalities are provided by specic solutions as e.g. TransSearch, see Bourdaillet et al.

7_{I will not delve into the question of whether localization truly denes a separate activity or designs}

(2010). A solution like Linguee makes use of the web as a parallel corpus, see Linguee (2011).

Translators use many other computer applications such as desktop searches, voice recog- nition software etc., which are not typical of the translation process. Therefore, although they are essential for everyday work, they are not listed here, but can be found e.g. in Austermühl (2001). A special note is necessary only for word processors because they are used as editors by some translation environment tools, see 1.4.3.