Translation is, or at least should be, a regulated and standardized process that results in a product, i.e. translation, of the highest quality. Although the quality of the target text depends on a number of different factors such as the quality of the source text, the deadline, the price, how focused the translator is etc., there are three basic steps – translation, revision, and proofreading – that should be taken in the translation process in order to always guarantee the satisfactory quality of the translation.

The first step, i.e. the translation, is a step that is implied and the step for which the client is always willing to pay. The following two steps – revision and proofreading – are just as important in the translation process. If these steps are skipped, there will almost certainly be errors in the target text that the translator has overlooked. Revision is the process of reviewing the text by comparing, or reading the original and the translation side-by-side. The person in charge of this step should have equal qualifications to the translator. In this way, the person reviewing the translation identifies errors made by the translator such as typos, wrong numbers, omitted words or sentence elements, word choice errors, and so on. This step is crucial for producing a high-quality translation. Without it, the client may end up with a translation of an invoice that says that they have to pay EUR 300 instead of EUR 300 thousand because the translator was not focused. The third, and often the final step, is proofreading. Depending on the source and target text, the task of proofreading, i.e. the final check of the translation without consulting the source text will be entrusted to a native speaker of the target language. Thus if the translation is, let’s say, from Serbian into English, proofreading will be done by a native speaker of English. Another reason why this step is so important is because even after the revision process, the target text might contain some syntactic and stylistic errors that the translator and the person in charge of revision simply fail to notice because their linguistic intuition is not on the same level as the intuition of a native speaker. If this step is omitted, the client will get an accurate and high-quality final product, yet it will probably not be 100% natural and adequate in terms of style.

In addition to these basic steps, there is another important segment to be considered in the process of checking the quality of the translation. If more than one translator is working on a large-scale project, in addition to revision and proofreading, it is also necessary that the translators agree on the terminology that they will be using, which is done either before or during the translation process. This step is necessary in order for the terminology and the style of the target text to be consistent and uniform after the translation process had been completed. If the translators are provided with a term base, they should stick to the existing terminology. In addition to accuracy, this guarantees terminological uniformity, which is crucial because translating the same concepts differently throughout the same text would cause confusion.

Finally, it should be noted that there are quality assurance tools that make this process not only easier, but also much more precise. These tools help identify inconsistencies, perform spellcheck, detect double spacing and formatting errors, etc. Such tools are very useful when one project is delegated to not only several translators, but also several people in charge of revision.

Therefore, quality assurance is just as important as the translation itself. It guarantees consistency, coherence, and accuracy of the target text, which is certainly what the client expects. With that in mind, it is clear that quality assurance, as a key element in the translation process, should never be omitted and disregarded, regardless of the length and complexity of the text.

As the name suggests, the Translation Memory system is a type of database that is used in computer programs designed to help translators.

Translation memories are typically used in conjunction with computer-assisted translation (CAT tools), programs for data processing, terminology management systems, and multilingual dictionaries. Translation memory consists of segments of text in the source language and their translations into one or more target languages​​. These segments can be passages, paragraphs, sentences or phrases. Individual words are not in the translation memory domains; they are processed in terminology bases.

The translation process with a TM system

The translator first enters the source text that needs to be translated into the translation memory. The program then scans the text in order to find segments in its database and displays them to the translator. The translator then examines the offered translation and can accept, reject or modify it, and then use the modified version. In this case, the modified version is saved and stored in the database. When using translation memory, varying degrees of concordance can be determined/used. Segments with no concordance matches must be manually translated. These new segments are stored in the database so that they can be used in other translations.

It is important to note that a translation memory is not a system for automatic translation. The system “does not understand” the text and it cannot translate itself, but can only store what the translator has entered. It serves only as an aid to the translator. The TM system is only one option/ way to store translations. It does not check whether there are linguistic or grammatical errors in the translation. If an error is stored together with the translation, it remains in the TM system until the translator detects and corrects it.

In addition, it can be used effectively only when the original texts have adequate quality. Apart from that, one must take into account the structure and consistency in the use of terminology.

Translation memories are most useful when translating texts with a lot of repetition, as is the case with technical documentation (user manuals, etc.) and documents containing specialized terminology. Translation Memory is not considered appropriate for literary texts because there are few repetitions.

Benefits:

• High quality translation
• Consistency of terminology – this is important when different translators are working on the same project
• Speeding up the overall translation process, shortening delivery times and thus
• Reducing the cost of translation projects.

Software Localization

While the translation of the majority of product-related supporting materials, such as documents, license agreements or marketing materials requires the use of translation memory systems or various terminology management tools, there are special localization tools for the localization of user interfaces of various software products.

Localization tools are special software applications that are based on translation memory technology and are designed to help the translator when processing texts of user interfaces (menus, dialogs, system messages) and adapting the graphic elements. One of the main features of software tools is their WYSIWYG (What You See Is What You Get) function. This means that the translator has the ability to see the current translation unit in its real context. On the other hand, the translator can adjust the text field to the length of the target text and edit images with text or culturally variable elements. During localization, the translatable text is extracted from the source code, and at the end, the translated text is entered back into the source code.

These tools have extensive functions for quality assurance and the possibility of integrating dictionaries/glossaries.

Software products can be functionally divided into a number of components. The most important components within software localization are the user interface and the documentation.

Most localization projects begin with a localization of the application’s graphical user interface. It is a central component of the localization. The following components of the user interface are to be localized: menus, dialogs, and strings (warning messages, error messages, status messages, etc.), as well as various sample files that are used for demonstration.

All the existing commands of a software are grouped into menus. A menu is a list of different commands, i.e. options. By clicking one of the menu options, the menu expands by opening downwards, so that it is also called a dropdown menu. In the dropdown menu, each option can have multiple sub-options, i.e. it represents a group of related commands. The context menu is obtained by clicking the right mouse button. The menu appears at the point where you clicked and contains commands that can be applied at that time.

The dialogues are a special type of window that poses questions, enables you to select an option for performing a task or provides information.

Hotkeys and shortcuts are the components that appear in menus and dialogues. A shortcut is a command on the keyboard which can be specifically called up. It is characterized by an underlined letter in the menu name, menu item or command, and is called up via a corresponding key combination.

Graphic elements: Icons are small graphic symbols which represent various programs, files, and other objects. In addition, they also serve to activate/give commands and run programs.

Documentation

The software may include: printed documentation that includes installation instructions and user manuals as well as marketing materials, registry cards, etc., and online documentation: online guides, web pages, and online help.

Online help is usually the most extensive component of the translation of most localization projects. The advantage of online help in relation to printed documentation is that the user can directly access help via a web page or user interface, without interrupting any work. This is done by selecting a specific topic from the Help menu, clicking on the link or button on the user interface, or by pressing a special key. Online help includes direct answers to specific questions and displays them in a separate window or in a separate dialog box. Texts in the online help system are divided into smaller sections/areas so that the user does not need much time to find the desired information. Online help structure follows the hypertext principle. This means that it is possible to jump from one section to another.

The development of information technologies has, among other things, automated the field of translation. The arrival of machine translation is, therefore, a logical consequence of the overall automation that has affected all spheres of life and all activities.

However, despite the initial optimism and enthusiasm reflected in the expectation that the further development of machine translation will bring relatively rapid and significant results, it has not yet reached a level of reliability that would provide completely automated translations.

Machine translation is the use of computer software for the purpose of translating text or speech from one language into another. Ideally, this process involves an analysis of the source text, and the target text generation without the need to engage people. The reality is, however, different. Machine translations rarely require no prior or subsequent processing.

The idea of ​​machine translation is as old as the first computer systems. The original machine translations from the mid-twentieth century were a simple substitution of words from one language with the words of another language, provided that the corresponding pairs of words were found in the preset database. Later models were not limited to words only, but used entire phrases, which significantly improved the grammatical structure of the text and translation of phrases where a literal translation was not possible.

Machine translation, as well as any other software application, requires extensive programming. This process involves creating a bilingual or multilingual termbase and entering grammar rules, grammar structures, and numerous examples based on which the system determines the parts of the target text corresponding to the parts of the original text.

Notwithstanding the significant improvements in machine translation, which is no longer a simple substitution of words from one language with the words of another language, and which even adopts and implements new rules based on the context, machine translation still has a very limited use. Its main significance is reflected in the assistance it provides to translators, whereby the results depend on the type of text.

The translation process and the challenges of machine translation

Translation as a process includes complex cognitive processes that involve interpretation of the source text and interpretation of meaning in the target language. Translation is almost unimaginable without active thinking because it requires an inexhaustible knowledge of the grammar, syntax, semantics, polysemy, idioms, slang, jargon, etc. of the source language and knowledge of the culture of its speakers. The same kind of knowledge is necessary for the interpretation of the meaning in the target language.

Therefore, the greatest challenge to machine translation is reflected in the need to program computer software to “think” and “understand”. Ideally, the software would be able to understand the text the way that people do and able to generate a target text that sounded as if it was not written by a machine. This, however, requires a higher level of technological development, especially in the area of artificial intelligence.

With this in mind, machine translation, at best, automates the easier part of the translation process. Everything else that requires research, solving ambiguity and identifying and processing errors, as well as complex cognitive processes remains unattainable for now.

Of course, machine translation, and work on its development are not useless. Already at this level, machine translation can have positive results. Some of the most popular tools for machine translation actively use the entire content available on the Internet that includes millions of translated documents, in order to find linguistic patterns and establish connections between different languages​​. The main problem is the fact that such results can be achieved only in the case of certain types of texts, using controlled language and final processing which, of course, is not done by the machine. Additionally, the content on the Internet is not error-free and elimination of the existing inadequate texts is often impossible.

Entrusting the entire translation process to a machine would omit the fact that communication takes place in a context that only people understand. In this way, language, which is primarily a social phenomenon, would be dehumanized, and the translation process would be reduced to mathematical and statistical operations, largely ignoring the linguistic and social components. Translating out of context, that is, literal translation is completely useless in many cases, and in fact may sound ridiculous, cause misunderstanding or, in worse cases, may be an insult to the speakers of a language who are also members of a particular culture, with unforeseeable consequences.

Despite the results, which are not impressive so far, the development of information technology will undoubtedly provide at least somewhat acceptable translations in the future. The question is to what extent and how far in the future. Developers and experts will improve artificial intelligence and create a vast knowledge, vocabulary and memory base, however, it is difficult to imagine computers using common sense, learning from experiences in the way that is typical of humans and having intuitive knowledge in the near future. Machines are still not able to “understand”, which leads to the conclusion that the future of machine translation is in providing an understandable basis which will further be improved by translators and linguists.

Machines versus humans

All this is a consequence of complexity on one hand, and the imperfection of the human brain and reasoning on the other hand. The difference in the mechanisms by which computers and the human brain operate has so far been insurmountable. The human way of thinking involves understanding, analytical thinking, and the application of experience.

Although it cannot reach the volume and speed with which computers have processed information for a long time now, the human brain is characterized by processes that, unlike computers, are complex and therefore still mysterious and unknown. Machines do not understand the concept of meaning, they only substitute one text with another based on information made available in advance by humans.

Translation without thinking and understanding is unthinkable, or at best, it can be good if a number of prerequisites are met or if it is a mere coincidence.

The biggest difference between translators and machine translation systems may be seen in the fact that people regularly and almost unconsciously adapt to all circumstances. Only humans have the ability to understand a deliberately or accidentally incomplete, imprecise and ambiguous text, and only humans can adequately adjust the text to the target audience, perform various compensations, recognize false friends, and recognize the lack of corresponding grammatical categories or terms, as well as errors, and decide how to treat them. Likewise, only humans can choose a certain synonym based on their knowledge, experience, habits, and personal style.

The interpretations of translators are based on various decisions, lifelong experience, and understanding of meaning, and all we can hope for from machines is a more or less successful simulation of this. Translation is a decision-making process, and machines often do not make the right decision, so the result of machine translation is often not something that could be considered natural language and communication.

Basically, machines are not smart, and they are insensitive as well. Their “knowledge” of grammar and vocabulary is limited to what the developers have managed to enter. All it takes to obtain a machine translation is to formulate commands and provide a term and phrase base.

Adjusting to machines in order to adjust them to humans

As already mentioned, there are still cases in which machine translation produces results. This primarily refers to controlled language, or texts adjusted to computer processing. Such texts would normally need to have relatively short sentences, simple structures and a simple message that is transmitted through the text. It is necessary to avoid ambiguity, metaphors, puns, slang, personal touch, idiolect, idiomatic expressions, expressions typical for dialects, sociolects, jargon, abbreviations, etc.

Pronouns can also be added to the list of things to be avoided due to word order and differences resulting from natural and grammatical gender. Clear, grammatically correct sentences should be used, without omitting any words. The adjustment of the date format and numeric values​​ is also of great importance, and it would also be necessary to include a list of terms that are not to be translated, because a machine does not differentiate between translatable and untranslatable content.

Having all these requirements in mind, it is clear that there are very few texts that meet the aforementioned and other prerequisites for a quality machine translation. Cases when this kind of text can be expected and when it is justified are quite rare. This is possibly the case with technical texts and reports containing many numeric values ​​with a limited number of established formulations, but beyond that, reducing a text to language “understandable” to a machine would be a violation of the language and a giant step back because it would lead to a simplification of language as the most perfect means of communication after several millennia of development and progress.

Machine translation of literary texts, whose structure often steps out of the established framework, and where it is necessary to preserve the atmosphere, nuances, and other finesses, is practically unimaginable as it is unimaginable for machines to write literary works of artistic value.

Benefits and advantages of machine translation

The list of deficiencies is considerable, however, machine translation certainly would not exist without some positive aspects and unimaginable potential. It can primarily serve as a starting point that will be further reworked, developed, and adjusted by translators and linguists. Machine translation is much cheaper, the translation is obtained almost instantly, the scope of the text has very little impact on the speed of translation, translation is always available regardless of time of the day, working hours, holidays, time zones, maximum productivity is achieved, etc. Another very important aspect is that these translations are confidential because they avoid intermediaries and their consistency is guaranteed.

The initial cost of investing in machine translation systems is relatively high, but, if justified, this investment is quickly rewarded due to substantial savings of time and resources.

Machine translation systems also provide a reliable text analysis, and therefore, the possibility of optimal planning and organization of work.

Machine translation is most useful when you need an urgent translation in order to gain general insight into a specific text, for example, when it is necessary to determine whether a particular term, and therefore a concept, idea or a message appears in a large volume text or in a large number of texts. This type of translation allows for the quick disposal of unnecessary material so as to focus on what is useful. Of course, all this is much easier if the text is intended for experts in a particular area who will easily gain this general insight.

The application of machine translation, and information technology in general, is multiple. The appearance of certain internet applications and services has made machine translation ​​available to a large number of people. Using free online translation tools like Google Translate, Bing, and BabelFish, millions texts are translated on a daily basis. Machine translation has allowed people to make contact via the Internet, various social networks, and communication systems that do not require a formal translation, but only communication of the general idea. At the same time, with the development of mobile devices, machine translation has become possible not just anytime, but anywhere, sometimes with unexpectedly good results. Mobile devices that offer the option of a machine translation, are becoming “more mobile” with time. One of the latest examples is Google glasses with a built-in translation tool – it is no longer necessary to even hold the device.

Potential clients from all parts of the world now have the possibility to instantly see the presentation and offer of a certain company, even when there is no website in a language they understand, and in some cases, websites even offer the option of translation in different languages ​​via integrated applications for machine translation.

This implementation of machine translation has also enabled traveling abroad without an escort translator, communicating without intermediaries, and finally, it has enabled the learning of foreign languages​​.

The essence of machine translation is in cost saving. However, in many cases, it may turn out that it would be less expensive to entrust the translation to experts and have it translated again. In addition, there is a big difference between free tools on one hand, and commercial or specially-developed tools for machine translation on the other. The best results will undoubtedly be provided by a tool that is specifically developed to meet the needs of a particular client, but in this case the price can be quite a problem for many people.

In order to achieve optimum results in the application of machine translation systems, it is necessary for translation services providers and their clients to be aware of all the possibilities, limitations, and expectations in advance.

The future of machine translation and the translation profession

Having considered everything machine translation can offer, as well as its numerous limitations, it appears that the future of machine translation is in the automatization of the part of the work performed by translators. Probably the most important application of machine translation is that of computer-assisted translation (CAT) tools.

The work of professional translators and the localization process is now virtually unimaginable without the use of these tools, that is, the functions they provide, including integrated dictionaries, the possibility of searching, replacing, recognition, searching for previously translated terms, analysis of the extent and characteristics of the text, terminology management, translation memory creation based on existing translations in order to facilitate future translation projects, etc. The advantages of these tools are numerous, and they are especially reflected in their guaranteed consistency and elimination of the translation of repetitious parts of the text, which leads to significant cost savings.

However, as long as computers are not able to think, machine translations will not be considered as reliable translations, and they are often blamed in order to avoid any responsibility for an inadequate translation.

One of the consequences of the development of machine translation technologies is the debates on what the future holds for translators. The fear of losing one’s job and position is somewhat justified. People are now replaced by machines in many fields, but as long as customers are not willing to trust a machine translation rather than a language expert, there will be a translation profession. Therefore, translators do not need to fear for their career more than the representatives of any other profession, because automatization has expanded in all spheres of activity.

Personal experience with machine translation

Considering all the advantages and disadvantages of machine translation, in this age when information technology, access to a computer, and the Internet have become available to most people, the best way to see for oneself the reliability of machine translation, or at least the model of machine translation which is free and generally available, is to conduct a personal experiment. All you need to do is enter the original text, for example, in a foreign language, in the appropriate field in the translation program and the translation will be generated instantly.

The knowledge of the given foreign language can certainly help to establish the extent to which the translation differs from the text in the foreign language, but even without the knowledge of the source language everyone is competent enough to assess whether the target text in their own language makes sense and whether it is idiomatic. Some experiments where a text was translated from a foreign language, and then returned to the original language, gave very interesting results.

Based on this, anyone can make their own objective judgment and form a personal opinion about machine translation and its purposefulness, and in some cases, it is possible to have fun reading an amusing translation of the text even when the source text is not at all funny.

The prevailing opinion on translation, as well as on any other profession with which many people do not come into regular contact, is usually based on a lack of information and on personal views. In most cases it turns out that this opinion is wrong, which leads to misunderstandings and disagreements when the need arises for using the services of this profession.

These misunderstandings can eventually cause the termination of cooperation leading to mutual loss: a translator or a translation agency loses a business opportunity, and the clients will often turn to an individual or company who will meet their requirements in accordance with their own beliefs, which often leads to poorer translation quality and continuation of erroneous opinions about translation. The objective of this text is to present some of the most common inaccurate perceptions about translation in order to explain its true nature.

Knowing two or more languages ​​automatically implies translation skills

This is probably the most common myth about translation resulting from insufficient knowledge of the translation profession and all the activities it covers. Translation is not just switching words from one language into another, and knowing more than one language does not guarantee that a person will be able to clearly convey the intended message and all its nuances. Most translators and interpreters are professionals with professional training and years of experience.

There is no difference between translation and interpreting

The term translator generally includes both translators and interpreters, so very few people know how different these two types of translation actually are. Translators and interpreters must master different sets of skills: translation requires extensive knowledge of the rules of language and translation tools, whereas interpreters must develop the ability to quickly understand and convey the speaker’s message relying on their own memory, resourcefulness, and the ability to take notes.

Machine translation will completely replace translators

Even though machine translation can be very helpful, its result still cannot compare to that of human translators, especially when it comes to longer texts. Translation must consider the context of the text, its slang, idioms, cultural nuances, and many other aspects which the machine translation is not able to process. It should also be noted that there is a tendency towards falsely equivalating machine translation (Google Translate) and CAT tools (e.g. Trados) which can sometimes lead to avoidance of using either of these tools and the benefits they provide.

Translation is always quick and simple

Many clients request translation services at the last possible moment, thinking that translation is a simple task that can be done quickly regardless of the length and the subject matter of the source text. However, the process of translation involves many steps including researching the topic and the subject matter, searching for correct terms, proofreading and editing the text, etc., where every step can take up a significant amount time so that the final translation can be done with precision and quality.

Any translator can translate any type of text

This myth arises directly from the previous one. Clients often think that translators are equally familiar with terminologies of numerous areas, regardless of how specialized those areas may be. Contrary to that belief, experienced translators are usually specialized in several related areas (e.g. media, telecommunications, IT) so that they can follow the latest trends and changes in those areas. Translators with years of experience in translating texts related to telecommunications can find themselves having a lot of difficulties when translating medicinal texts. For this reason, it is best to look for translators that specialize in the area that the client needs.

(more…)