Computer-assisted translation is a translation performed by a person with the help of a software application. The difference between computer-assisted translation (CAT) and automatic translation (AT) is that in the former the translator is involved at all time: it is the translator who translates, the machine is only there as an "assistant".

The CAT programs used by professional translation service providers generally include a text editor, a translation memory manager, a terminology manager, a project manager and automatic translation functionalities. Some of the most commonly used computer-assisted translation programs are Trados Studio, MemoQ, Transit, Wordfast and Déjà Vu.

In general terms, CAT tools are based on the reuse of texts already translated and revised by a person (stored in what is known as a translation memory) and on a system of "matches", making it possible to detect and leverage identical or similar sentences that have been translated before. When the program detects one such sentence, it prompts the translator, who ultimately decides whether or not to use it.

CAT programs guarantee a high degree of consistency between translations for the same client, as well as more competitive prices in the case of highly-repetitive texts (manuals, management reports, audits, working procedures, updates, etc.) through the application of what we call discount tables.

Automatic Translation (AT) or Machine Translation (MT) is carried out in its entirety by a machine. In these cases, the translator becomes the post-editor, and their function is to check the machine-generated translation. AT is normally based on rules or language corpora (or a combination of both) programmed in advance and configured by a computational linguist.

This type of translation is advancing in leaps and bounds due to the need for instant translation generated by browsing on the Internet, where free on-line translators abound. However, at this moment in time they only work well in highly-specific fields (generally speaking technical) and with language pairs from the same family. For example, Spanish-Catalan or Dutch-German automatic translators work relatively well, whereas Chinese-Russian automatic translators still have some way to go.

You should also take into account what you are going to use the automatic translator for: to get the gist of what is being said on a website? No problem in using it. To send technical specifications to a new foreign client? Totally inadvisable.

Created in 2001, Google Translate is a free on-line translator that can render translations between 64 different languages, which naturally include Spanish, Catalan and German. It is calculated that some 200 million people use it every month and, as occurs with other online translators, it leverages enormous databases containing multilingual documents and a statistical search system to locate the best translation.

Google's databases increase as users use the system, which means that all texts and documents translated by Google Translate remain on the Internet and can be accessed by any users of the system. The utter lack of confidentiality is one factor that should be borne in mind when deciding when to use Google Translate or not.

In terms of quality, the technical sophistication of Google's system is improving all the time, and the results are increasingly better, depending on the topic and language pair. Nevertheless, as yet Google Translate cannot produce error-free texts or render the necessary nuances or style depending on who the text is intended for.

For example, if we choose to install the Google Translate widget on our website so that when a German client clicks the German banner the page will be automatically translated, in all probability the text will contain errors, gaps and inaccuracies, with the corresponding lack of understanding by the potential client, thus jeopardising or compromising our company's image. If we use Google Translate to translate an advertising campaign originally intended for Spanish consumers automatically into German and launch it on the German market without any kind of adaptation, the results may be utterly and directly catastrophic. To say nothing of humanistic or literary texts.

Nowadays, the best way of knowing if Google Translate is suitable for our purpose is to use it first and then have a native speaker give their opinion on the translation generated by the machine. Considering the speed at which the state of the art is increasing, this might not be necessary in the not too far-off future, although for the moment, automatic translation results are still highly uncertain.

A translation memory (TM) is a bilingual archive or file comprised of original texts and their translations, in which the units (normally phrases or sentences, separated by a full stop/period) are aligned in order to form pairs between the two languages. We could say that a translation memory is a kind of bilingual database.

Translation memories are created from translations rendered with the help of a CAT program or else through the alignment of original texts and their translations using alignment programs. In this context, "align" means assigning the German translation to a Spanish sentence, or in other words, establishing a correspondence between the Spanish sentence  and the German sentence.

To have a better understanding of this concept, imagine an Excel glossary with two columns: Spanish and German. Instead of terms, this glossary may be comprised of sentences, i.e., one column with a Spanish sentence in each cell and a column next to it with the German equivalent (translation) of each sentence. We then plot a line between both columns, cell by cell, and "align" the original and the translation. Alignment programs perform this operation automatically and create translation memories in formats that can subsequently be opened with an assisted-translation program.

Computer-assisted translation programs use translation memories that can leverage previous translations and optimise both the consistency of the text and translation speed.

CAT programs have an analysis function that can compare the new text to texts already translated and which are contained in the translation memory. Once a comparison (analysis) has been made, the system generates a statistical output of the degree of matches between the new material and the material stored in the memory.

Thus, for example, it may tell us that 36% of the new sentences have already been translated and are stored in the translation memory. These matches are known as 100% repetitions, meaning that the sentence that appears in the new text has already been translated and that the translation is available in the translation memory. This often occurs when, for example, we update technical texts (manuals, software, data sheets, etc.), since in these updates only parts of the text, not all of it, are new.

This analysis function can also detect partial matches, i.e., sentences that are not 100% identical but have several or many elements in common. This type of partial matches are known technically as "fuzzies" and indicate matches of between 50% and 99%.

To account for these repetitions in the translation price, many translation service providers use discount tables according to the repetition percentages. Thus, and by way of example, a discount of 30% may be offered on the original price for 100% repetitions — let us not forget that they must always be read, checked and then included in the text, which involves work that must be charged for —and a discount of 15% for repetitions of 80% to 99%.

The most widespread translation standard nowadays is the UNE-EN 15038 "Translation services - Service requirements", which regulates all the procedures involved in the rendering of a translation service. Since translation is a product the quality of which depends both on objective (suitability, correct spelling) and subjective (whether or not the style is pleasing) criteria, this standard does not seek to establish a finished product quality control but rather a quality control that spans the entire production process (project management, selection of a suitable provider, profile of the professional translator, revision, post-editing, invoicing, claim or complaint management, etc.).

As occurs in many other industrial and professional areas, the ISO 9001:2008 standard is very widespread and determines the requirements for a quality management system. Although this standard is general, and is not specific to translation, its implementation makes it possible to establish quality control mechanisms.

At this moment in time, the ISO/TC 37 Technical Committee for Translation, interpreting and related technology is working on the transposition of the UNE-EN 150382 into an international ISO standard, as well as on the drafting of standards for interpreting, post-editing, automatic translation, simultaneous interpreting equipment and technology, to name but some.

ProZ.com is an international translation portal where supply and demand converge. Translation providers (companies and freelancers) post their CVs and credentials and clients who need translations try to locate the most suitable provider. Clients can also post their offer and award it to the "best bidder", popularly known as a "translation auction", a practice whose results are not generally very recommendable in terms of translation quality. You get what you pay for.

ProZ.com's basic services are free, although they are very limited. Access to full functionality means becoming a premium member and paying a fee. Apart from its job vacancies, ProZ.com's other services include linguistic fora, glossaries, news, references and training.

The advantages offered by ProZ.com to clients include immediate access to numerous professionals of all languages who are willing to compete with each other to be awarded the commission, with the subsequent price advantage that this may involve. Some of the disadvantages are that the portal's administrators do not control the CVs that are posted, either with regard to the degree of professionality or truthfulness.

There are other translation portals, such as TranslatorsCafe, Ediciona, LanguajeJobs or TranslatorsVillage.

From the strictly linguistic standpoint, Latin American Spanish (LA Spanish) as such does not exist, because the varieties of Spanish spoken in Latin America are practically the same as the number of countries in the continent. Thus, the Spanish spoken in Argentina and the Spanish spoken in Mexico are not the same, as neither is the Spanish spoken in Venezuela or Uruguay. The name is therefore a general and arbitrary one that is used to distinguish between the Spanish spoken in Spain and the Spanish spoken in Latin America.

There is a belief that the origin of this expression goes back to when the Microsoft Word spellchecker decided to create multiple variants of Spanish, including all the Latin American countries, and all the "Traditional Sort" (the language of Cervantes?), and "International Sort" (the Spanglish of New York?) variants. 

Linguistic purism apart, there is no doubt that in the translation sector this "variant" of Spanish is here to stay, and nowadays it is not uncommon to receive requests for translations from German into Latin American Spanish, particularly in advertising and marketing. This merely means that the product is intended for a Latin American audience, and that the Spanish version must take this requirement into account. Of course, a text created in LA Spanish will never convey the language and regional nuances or differences between all the Spanish-speaking countries, although it can have a style, turns of phrase and terminology that are better adapted to Latin American than Spanish usage. There are numerous differences, which is why an adaptation to LA is necessary.

Multilingual navigation on the Internet has led to the necessity to create immediate content translation systems in order to obtain information, buy, book, publish, communicate, etc. This is why there is now a proliferation of free online web-based translators.

As a rule, these online translators are automatic translation systems based on multi-language databases comprised of hundreds of millions of documents (see a representative example of German/Spanish here). When we use an online translator to translate a sentence, a text or a website, the translator will run a statistical search in its huge database in order to detect total or partial matches and will use them for the translation. This means that the machine does not think about how a human being would do it and neither does it perform the translation based on grammar, meaning, syntax or context: it translates solely on the basis of statistical calculations of matches in texts that have already been translated. The quality of the matching texts located in its database will determine the quality of the final translation, together with the online translator's degree of technological sophistication.

It is very important to bear in mind that when we use a free automatic translator, all the texts that we input into it or have it translate will become part of the online translator's database, which means that it is impossible to guarantee the confidentiality of our texts. Getting things free does have its downsides on the Internet. The language corpora of online translators are fed with the users' documents, so by using this kind of translator we are giving our express consent for the information provided to be used in any way and for any purpose, thus losing our control over it.

The most common price unit used to prepare a translation quotation is the source word. This unit is being used increasingly more on the market due to the use of computer-assisted translation (CAT) programs, which use the word as a basis for analysing the volume of text and the degree of internal repetition.

However, there are other units that are equally valid depending on the type of translation service. Thus, for example, revisions are normally quoted by hours or by pages, and in some countries, such as Germany, the line is the most common price unit (although as we have already said, the word is becoming more common in preparing quotations on account of the widespread use of CAT tools).

When we talk about "line" we are not referring to the line you might see on a paper or on the computer screen, or to a standard line of 8, 9 or 10 words, as is often used in translation quotations (a rather inaccurate calculation that does not reflect the real volume of the text). When we talk about "lines" we are referring to a standard line that is normally comprised of 50, 53 or 55 characters, depending on the country.

For further information about how to draw up a translation quotation for the German-Spanish combination, contact us at info@tradeus.eu.

The term “transcreation” is a merge of the English terms “translation” and “creation”. It is also referred to as “creative translation”.

It is currently a very fashionable concept in the translation industry, and it relates directly to one of the oldest maxims of the translation arts; that a translation should never seem to be a translation.

However, in this case, transcreation goes one step further and sets out to reformulate, virtually in its entirety, the original text in the target language in order to arouse the same effect in the readers of the translation that the original work caused among the readers of the original.

Naturally, this technique is used mainly in advertising, where besides the key selling concept, conveying emotions is a core requirement. For this purpose, transcreators freely adapt the style, forms of address and interaction, cultural references, names, places and any element of the original text that does not fit into the target culture for the intended purpose.

It is an extremely creative translation which, whenever possible, should be performed in close collaboration with the company or the customer's advertising department.