Machine translation programs
In order to understand which way the wind blows I searched the Internet and surveyed subscribers of the translation forums. The situation turned out to be threating. The issue is that they call TRADOS one of the representatives of an almost unknown type of the program package: Translation Memory Tools (TM-tools). If Russian users have already known text editors, browsers, mailing systems and desktop publishing, but when they are asked about computer tools for a translator, they all start to talk up or scold famous machine translation software (MT-programs). The difference between MT-programs sold here and popular TM-products in the West is almost the same as between astrology and astronomy.
General public treats machine translation programs as a universal mean, which is about - next Monday - to replace a human completely. However, specialists (in particular, developers of such programs) understand that this is far from true. Developers often sincerely write in their advertising literature: certainly, a computer translation is far from perfect, but it is better to have a bad translation than nothing. They do not deny that in order to get a “legible” text a human interference is required; they do not go public about the fact that such a human should be a qualified translator and should spend a lot of time to shape such a machine text.
As a result of the text processing by a machine translation program you get a set of words giving a general picture about the contents of the documents, but it does not convey the meaning fully, it is even not a text. You can easily think of activities which can accept such “translations”. For example, meteorological centers of different countries can use them for data exchange: vocabulary of meteorological messages is quite small, sentence structure is uniform, and if you do not use poets to make forecastings, an electronic translator will be able to adequately send the information after a respective setting. This is also valid for patents and other related fields. But a computer alone will not be able to cope with most of the texts. Herewith mistakes made by such a “brainless” assistant are so unexpected and numerous, thus professional translators will not spend time on its training.
TM-products are totally different type of programs. They are based on the use of the translation memory. Translation memory is a special database created by a translator during work. As well as an electronic dictionary this database contains pairs of terms in two languages, but unlike an ordinary vocabulary, it uses a whole phrase as an elementary unit instead of a separate word (up to several lines of the text). Such memory is formed based on the files already translated by the user: i.e., work with TM-tools often starts with uploading of a pair of texts - an original text and its translation, used as the basis for the memory. It is clear that the larger an original pair (or several pairs) is and the more qualified the translation is, the more useful the memory will be.
When using the memory, a translator, as a rule, uses three windows. The first window contains the original text, the second is used for a translation to be entered, the third shows hints which appear in progress. Let's assume you meet phrase “a customized parameter” you have already used last time. Now the hint window shows this - ready- translation, and you can agree to use it again with one click. Certainly, if you translate a Shakespeare, a number of such repetitions will be small.
But when translating computer documentation, you will meet most of the phrases more than dozens of times. Herewith you can set a TM-program so that it provides hints not only in case of a complete match, but also in case of a partial match (you may specify a match percentage). Firstly, it will help you cope with endings change typical for Russian language (as except for phrase “a customized parameter” you can meet phrase “under a customized parameter”, or “parameters customized by a user” and so on). Secondly, as the translation of phrase “Choose clause New from menu File” resembles the translation of phrase “Choose clause Font from menu Format”, then after setting the program to search not only exact matches, but close phrases, you can use the same template for translation of all phrases of this kind.
Thus, setting of a certain match percentage will allow you find an equivalent for many similar phrases with the help of one element from the memory. With progression through the original file your memory will be filled up, though for some TM-products in case certain settings are used, acquired knowledge is used at once for translation of the following phrases. In other words, if you once thought how to translate phrase “next follow the instructions on the screen”, then all repetitions of this phrase you will meet will be automatically translated that way. A friend of mine told me that at some instant the text he was working on, turned to be completely translated as all next fragments have been met before.
So, I will summarize the fields of TM tools use and its advantages. Translation memories are useful when working on large uniform materials with a great number of repetitions. Especially when it concerns multivolume tomes translated by large teams. In this case a translation memory can be originally made up by a leading translator of the team and then shared with other team members. Except for a significant energy saving, it will ensure the unity of terminology which is so hard to achieve when even two persons working together. Average efficiency improving, according to the users, makes 30 per cents, but for some texts the speed of translation may increase by several times.
Besides a multiple use of fragments once translated, most of TM-products allow you automatically save the formatting of the original text in the translation without delving into the finer points of an editor. The secret is that the text originally submitted with PageMaker, FrameMaker or QuarkXPress, is reflected in the window of a service TM program editor you use. Thus, your task is to replace the text in one language with the text in another, herewith the formatting will be saved automatically. As a result, you can work with a text using any editor (for which you have only a product service editor).
It is obvious that for an efficient use of TM-program translated texts should be submitted as files (if you need to scan and recognize texts printed on paper before using a translation memory, you can hardly save a lot of time).
|Deja vu||Atril||It is a low-budget and the most multifunctional product. Translators using it should apply for a certain useful function to be included in it, and such applications are executed in a timely manner.|
|Translation Manager||IBM||According to specialists this is the most effective though not a user-friendly program.|
|SDLX||SDL||A newly introduced product which is improving rapidly.|
|Transit||Star||It is most suitable to work with Cyrillic alphabet. An important advantage is presence of a Russian distributor.|
|Translation's Workbench||Trados||This program uses MS Word as a service editor. Considering that Microsoft purchased the shares from Trados Company and plan to use its software as a localized components internal store, Translations' Workbench program (simply called Trados) has good changes to become an international standard.|
Certainly, translation memories have disadvantages. Besides the limitations on translated texts mentioned before, it is a high price on most of TM-products (one thousand US dollars and more) and lack of the standard. Increasingly inviting translators for work foreing clients specify that an applicant must have a certain TM-program. However, you can find different titles in the advertisements, thus, when purchasing a product focused on one employer, you may not meet the requirements of another. And you need to spend as money as time to learn each product.
When estimating the costs of these programs you need to keep in mind that translation packages are sold as a set of modules (utilities for creating and maintaining a database, an editor, filters for working with various word processors, etc.), you may not need a part of them. Thus, you should pay attention what is included in the standard package and what is subject to an additional fee.
Hardware appetites of the considered product class are quite predictable (the more, the better - as usual). Moreover, on top of that, TM-products place increased requirements to the screen. As friends of a translator are multiwindow programs, you should better get a large screen to work with them. Specialists recommend to have at least a 17 inches screen.
I am not supposed to compare products of different companies - I just wanted to pay the readers attention to a new software layer. If you want to learn the specificity of various programs, you may visit the websites of leading TM-product developers.
I can help mentioning one more product - WordFisher. It is not widespread, and I have never seen the clients to require it. However, this program, developed by Hungarian translator Tibor Környei, has two significant advantages: simplicity and low price. WordFisher is a set of micro commands for translation of documents with MS Word and distributed as a shareware at a price of just 30 USD.
And the last “thought-provoking information”
The survey performed by the International Institute of Translators showed than now only about 15% of translator use translation memories for their work; however, a share of well-paid translators who use TM-products makes 40-50%.