Translation white paper

With the use of Internet and the development of electronic businesses, the need to comunicate quickly in all languages has become a must, thus translation softwares have become essential.

This white paper will allow you to better understand the way a translation software works and what you can do with it.

Contents
To find an answer to your question directly, click on the corresponding chapter.

1- What is a translation software?
   1.1- Definitions
   1.2- Applications
   1.3- Components of a translation software

2- Answers to the most frequently asqued questions (FAQ).
   2.1- What is the difference between a translation software and a translation dictionary or a translation memory?
   2.2 - Does Reverso translate literally?
   2.3 - Does Reverso generate a perfect translation?
   2.4 - Is it possible to connect dictionaries on CD-ROM (Collins …) to Reverso?
   2.5 - What are the various sources that were used to develop Reverso (grammars, dictionaries, etc.)?
   2.6 - Can Reverso replace a human translator?
   2.7 - A good opportunity for the translators.
   2.8 - - How can you translate a text on paper ?
           Is it possible to use an OCR?
   2.9 - Is it possible to translate pdf files or text contained in an image?

1. What is a translation software?

1.1 Definition

A translation software is a computer application that generates automatically a translation of any type of texts (letters, reports, articles, Web sites …) from a source language to a target language. The translated text must be as correct as possible in the target language and give back the information contained in the text of origin.
This tool can work in PC, Mac, Internet or Intranet environment, either as an autonomous application (word processing for example) or as translation functions directly integrated in an application (Intranet, Internet, Word or Excel …).

Reverso is available for:
PC (Windows 95, 98, NT …), Mac (OS 8 , 9, X clasic version), Intranet (Windows NT and soon UNIX server …), Internet (Internet Explorer, Netscape Communicator).

You can test the Reverso translation on these sites:

Softissimo : www.reverso.net, www.translate-online.com

There are other tools that can help you with your translation and which add up to a translation software, but which do not generate the same results: translation memories (like Trados) and translation dictionaries or electronic dictionaries (bilingual like Collins or multilingual like Eurodictionary).

1.2 Applications

Translation softwares allow you to:

Understand in real time a document drafted in a foreign language,

Translate faster and more effectively.

Help for the understanding
With the development of Internet, E-mail and Intranet systems in companies, users now have an urgent need to understand at once information in various foreign languages.
About 30 % of all the Web sites are not in English but in Spanish, French, German... Access to all the information is only possible when one knows several languages.
Therefore you can find great value in a translation software that will allow you not only to understand a text written in a language you do not master, but also to understand it immediately without the help of anyone.
A translation software gives you an autonomy that no other solution can offer.

Productivity tool
With the internationalization of companies, and the growing importance of electronic business, most companies have to diffuse information in several languages (internally as well as externally). Employees often have to work on information in a foreign language and therefore to translate documents before using them.
But the translation departments in companies are often too busy to satisfy everyone.
By using a translation software, the human translators can accelerate the translation process, spare a lot of time on the first stage of the translation process, and concentrate on the exactness of the chosen words. The human translator can become a true specialist of a specific subject , because he can focus more on the effort of localization than on the translation itself.

Some examples:
1- To translate a contract, the translator should be a jurist and know the Law of the two countries, he must localize the translation according to the Law of both countries.
2- In the business world: in the United States, there is no exact equivalent for the function of the French “PDG”. It is necessary to know the exact function to be able to choose the right translation : President, Chairman or CEO. Here again it is up to the human translator to choose the right translation and not to the computer.

Who can use a translation software?
Here are some types of users who might someday need a translation software. You will probably recognize yourself in one of these descriptions:

You are English-speaking, you do not speak foreign languages and you wish to understand Internet sites, newspapers, or other documents written in these languages, and / or you wish to send e-mails or letters in another language.
You speak French or Spanish, you do not speak English but you have to communicate with English-speaking people.
You master two languages and you have to translate a lot of texts (translators, bilingual secretaries), you wish to speed up the translation process (the use of the software can make you gain up to 40 % productivity).
You work in a French subsidiary of a German, English or American company, you need to understand all the documents inside the company : instructions, manuals, reports, brochures, internal notes.
You work in an international institution in an English-speaking country (embassies, representations of the EEC, OECD, UNICEF) or, the other way around , in a French delegation abroad.
And more generally, you have to communicate once in a while in a foreign language or to do some research : researchers, academics, students, journalists...

Return on investment with Reverso
Reverso allows you to gain from 30 to 60 % of the time dedicated to the translation process, so it can be made profitable after a few translations.

The more you use a translation software by personalizing it (by adding your own words or by connecting your dictionaries), the better the generated translation will serve your needs. The translation software is the only solution which makes it possible for you to understand immediately a text written in a language unknown to you. And that is priceless !

1.3 Components of a translation software

A translation software is made up of:

Linguistic rules
Dictionaries: developed specifically for the translation software.
An interface: according to the products, this interface allows to preserve the layout, to revise the translation, to adapt and personalize dictionaries …

1.3.1- Linguistic rules

The translation engine is the technological core of the translation software. The stages of the automatic translation process are:
- Analysis of the source text.
- Application of the linguistic rules of transformation and of desambiguation.
- Generation of the translation.

Here are a few examples of the applied linguistic rules : the order of the words, the syntactical and grammatical information, the morphology. The analysis clarifies all the elements of the sentence.

The finer the rules, the more precise the translation!

The Reverso translation engine

The Reverso translation engine analyzes the context: it does not translate literally.
Reverso combines the use of the most modern computer technologies (architecture 32 bits, programming object …) and more than 10 years of experience.
Reverso’s translation engine and its associated dictionaries allow to analyze and generate several hundred thousand forms (words, expressions …) offering a unequalled quality of automatic translation. And the translation is extremely fast: 1 page per second!

Here are some examples of linguistic forms handled by Reverso:
- Category of word: name, verb, adjective, adverb.
- Morphological information: conjugation, flexion.
- Semantic information: allows to specify a category for names (person, geography, firms …).
- Syntactic information : information about the construction of verbs, names and adjectives ( prepositional, infinitive constructions...).
- Use of an article.
- Order of the words.

1.3.2- Dictionaries

The dictionaries integrated in the translation software are not only a list of words or expressions with their translation : every word or expression must be defined with linguistic information (morphology, semantics, syntax …) in the source language and then in the target language. This information is then handled by the translation engine.

The richer the dictionaries, the better the translation.

Reverso dictionaries
Languages: French, English, German and Spanish.

General dictionaries:
The dictionaries supplied with Reverso administer more than 1 million forms (words, terms, expressions, declensions …) for every translation direction.
These dictionaries cover as many fields of activity as possible (computer, law, business, medicine, current …) to provide the user with a first translation as satisfying as possible.

Specialized dictionaries:
The specialized dictionaries are intended for specific translations (computer, medical, motorcar, petroleum, serviceman …). Every specialized dictionary is co-developed with linguists, lexicographers, translators and specialists of the field in question.
The work of lexicography, which has to be done for each dictionary, is based on the lexicon, the expressions corresponding to the domain, the names of companies, as well as the abbreviations or the acronyms.
The dictionary is then validated twice :
- In context with specialized corpuses,
- By an expert of the field (professor of medicine, scientist, computer specialist …) with the help of a linguist.

1.3.3 - Interface

The translation interface of the software is very important, because it facilitates the generation and the use of the translation.
Once the translation is generated by the software, the user needs to revise it, personalize it to his / her style, adapt it to his / her specific needs … If the aim of the translation is simply understanding, there is no need to revise the document, but if the document is to be braodcasted, it must be corrected first.

The interface of a translation software consists of:
- Various translation modules
- Translation and revision tools which make the translation and revision process of the text easier.
- Personalization tools which allow to create dictionaries by integrating terms, expressions...

The modules and the other tools available depend on the version of the software and the chosen environment.

Modules of translation:

Presentation of the source text and its translation in two windows.
Conservation of the pagination.
Integration in Microsoft Word, Excel (95 , 98 , 2000), Internet Explorer (4 , 5).
Translation of several documents at the same time (without opening them).
Translation of Web sites.

Translation and revision tools:

Adaptation of the source and target paragraphs.
Listing and marking of the unknown words.
Creation and marking of preserved words (words or sentences that are not to be translated).
Alternatives of translation when the context is ambiguous.
Translation of a word in a bubble.
Recording of the bilingual text.
Possibility of translation paragraph by paragraph.

Tools of personalization:

Creation of dictionaries and updating of existing dictionaries,

Encoding of the vocabulary according to your level of linguistic knowledge (beginner or expert). Connection of specialized dictionaries.

2. Answers to FAQ.

2.1 - What is the difference between a translation software and a dictionary or a translation memory?

Translation dictionaries
These electronic dictionaries are generally CD-ROM versions of known paper dictionaries. They make it possible to find the translation of a word or an expression according to the context, but they do not translate sentences. They add up to a translation software. Softissimo also publishes electronic dictionaries:
Collins : first-rate French-English / English-French bilingual dictionary Collins on CD-ROM (also supplied with the English Collins dictionary and the thesaurus ( alphabetical documentary collection).
EuroDico : the mutilingual electronic dictionary in 6 European languages: English, French, German, Spanish, Italian, Dutch.

Translation memories
Translation memories on a data base allow to store sentences or already translated paragraphs and to recognize them during a new translation. To use a translation memory, it is necessary to enter every source sentence and its translation in the base. Before being able to use it, a long and expensive work is necessary, which can usually only be achieved after a few years (according to the volume and the type of documents). Translation memories are especially useful to people who translate a large amount of documents very frequently.

2.2 - Does Reverso translate literally?

No. But let’s have a look at a few examples.
Some sentences below were translated with Reverso.
These examples will show you that Reverso does not translate literally, but that it analyzes the context to apply the right rules of transformation according to the language.

French - English
   1- Verbal expression ("mettre au courant de") and passive relative construction:

      Il n'a pas mis ses amis au courant de l'offre spéciale qui lui avait été proposée.
      He did not tell his friends about the special offer which had been proposed to him.

   2- Verbal expressions ("être à son compte", "avoir afaire à") and adverbial phrase:

      Je suis à mon compte depuis une semaine et j'ai souvent affaire à des fournisseurs douteux.
      I have my own business for a week and I often have to deal with doubtful suppliers.

   3- Expression ("bien vouloir" + infinitive verb), idiomatic expression ("sous peu" = "bientôt"):

      Nous vous remercions de bien vouloir patienter, votre docteur arrivera sous peu.
      We thank you for waiting, your doctor will arrive soon.

   4- Construction "soit..., soit...":

      Soit vous acceptez cette offre, soit vous faites une meilleure proposition.
      Either you accept this offer, or you make a better proposition.

English - French
   1-In the following two examples, the relative proposition is perfectly translated in French, even though there is no relative pronoun (whom, that…) in the English sentence:

      The person I talked to is right here.
      La personne à qui j'ai parlé est juste ici. (and not: la personne j'ai parlé à est droit ici)

      I'm impressed by the software I was shown.
      Je suis impressionné par le logiciel que l'on m'a montré.

   2- Conditional proposition with "Should":

      Should you need a car, please let me know !
      Si vous avez besoin d'une voiture, s'il vous plaît faites-le moi savoir !

   3- Transformation: intransitive verb -> pronominal verb:

      We shook hands.
      Nous nous sommes serrés la main.

German - English

   1- Expressions like « entweder, oder… », « ansonsten »… are recognized by the software:

      Entweder arbeite ich oder ich schlafe
      Either I work or I sleep

      Ansonsten lese ich
      Apart from that I read

   2- Whole structures are recogniezed:

      Du verhältst dich, als ob ich blöd wäre
      You behave as if I was silly

      Du verhältst dich, als wäre ich blöd
      You behave as if I was silly

2.3 - Does Reverso give a perfect translation?

The first goal of machine translation is to understand and / or to be understood. Therefore, the opinion of the user is the main criterion to judge the quality of the product : is the translated document usable and does it give the contents of the source text back?
This answer depends naturally on the knowledge of the reader and the supplied efforts, on the context and the nature of the source text, as well as the translation engine itself.

1/ Nature of the source text.
Reverso handles best texts without spelling or punctuation mistakes, written with a direct speech (neither literary, nor journalistic …), with a vocabulary of the current life and the business world.
To obtain the best possible translation, the original texts must be as simple and clear as possible. Avoid ambiguous structures. Choose short sentences instead of long and complex ones.

2/ Correction of the source text.
You have to make sure the text to be translated is correct in the first place:

Correct the grammar and spelling mistakes.

Respect the typographic rules : apostrophes must be attached to the following word, the right punctuation must be respected (commas, apostrophes, dots…) as well as the accents in the languages that use them (French, Spanish…).

These few pieces of advice will allow you not to get too many unknown words which the translation software can not analyze and which can sometimes change the meaning of the sentence.
The correction of the spelling mistake in a sentence is often enough to obtain a very good translation.

Example :
   We shook hand      Nous avons secoué la main.
   We shook hands    Nous nous sommes serrés la main.
Here: the mere omission of an "s" was enough to make the translation completely false!

3/ A translation software analyzes, but does not “understand”.
A translation software can analyze the structure of a sentence, the grammatical context, and transpose them into another language, without however “understanding” the context the way a human being would.
Very often, the linguistic rules can raise all the ambiguities:

Example :
The French word “comprendre” can either be translated by "understand" or by "include", but "comprendre que" will always be translated by "understand that".

But everything that is based on human intelligence (human sensibility, understanding of the world …) cannot be analyzed by the software.

Because the software cannot understand, there are cases when the generated translation will be bad and it will be up to the user to rectify the meaning of the text. There are other cases when the software will offer translation alternatives.

Let’s have a look at a few examples where the software cannot “guess” the meaning:

   1- Here are two examples in which only the reader can understand the meaning of the text:

      La jeune fille a aimé cet avocat.
      The girl loved this lawyer{*avocado*}.

      the trials lasted a month
      Les procès{*essais*} ont duré un mois

   2- Here is an example of a word which can have several completely different meanings:

      J'ai oublié les cartes.
      I forgot the cards.

Without any context, Reverso chooses the most current translation.
If you clarify the meaning, Reverso will choose the right translation:

      J'ai oublié les cartes de crédit. Peux-tu payer pour moi ?
      I forgot the credit cards. Can you pay for me?

      J'ai oublié les cartes routières, nous sommes perdus !
      I forgot the roadmaps, we are lost!

      J'ai oublié la carte des vins pour que vous puissiez commander.
      I forgot the wine list so that you can order.

2.4 - Is it possible to use dictionaries on CD-ROM (Collins …) with Reverso
         Or to include one’s own lexical bases?

Reverso contains a very rich general dictionary, which covers numerous fields of activity to meet the most common needs. Optional specialized dictionaries (computer, medical, motorcar, petroleum, military …) allow to obtain a more precise translation in very specific domains.

You can refine your translation by using classic dictionaries (Collins for example). With Reverso Pro, you can also create semi-automatically your own dictionaries.

2.5 - What are the various sources that were used to develop Reverso
         (Grammars, dictionaries...)?

The dictionaries used with the translation software are different in their structure from the classic bilingual dictionaries (choice of the translation according to the context).
The teams of linguists who developed the translation software worked with the best dictionaries and the best grammars, but they also defined rules according to the translation engine (see chapter 1.3).

2.6 - Can Reverso replace a human translator?

If you wish to publish a document or to establish a contract, the machine translation will not be enough : the revision of a specialist will be necessary for the translation of the complete document.

On the other hand, the machine translation can give you translations you would otherwise never have done ( you would never hire a translator to understand texts while you surf on the Internet for example, or to translate e-mails, etc.).

2.7 - An opportunity for the translators.

The translation softwares allow the translators to save a lot of time. Here are some examples:

Fast sweeping of the text to identify unknown words and relevant terminological expressions.
Direct recuperation of perfectly translated sentences and possibility to have simultaneously the source text and the translated text.
The translated text has the same layout; the layout is preserved when a modification is made.
Building-up of a terminology and guarantee of terminological coherence.
Possibility of offering new services with added value…

2.8 - How can you translate a text on paper?
Can one use an OCR (system of optical character recognition)?

You can scan your documents with a scanner and use an OCR software which transforms a scanned image into a text file. Most of the scanners are now provided with this system. Ask your computer retailer.

You can use an OCR system installed on your computer to use Reverso Pro.
The OCR system is not supplied with Reverso Pro.

2.9 - Is it possible to translate pdf files or text contained
in an image?

A translation software handles only electronic text. It can not translate text contained in an image.
Most of the files in pdf format are images (a pdf transforms a text into an image to protect it against possible alterations), therefore a translation software can not translate pdf files.
New versions allow to copy / paste the text : click on the text icon of your Acrobat Reader software , then select the text part you wish to translate and copy the text in your translation software, so it can translate it.

For all other versions, it is better to use an OCR software first (see above).