Practical tips to improve translation quality

The quality of translation

The quality of translation depends on the subject and style of the original text, as well as grammatical, syntactic and lexical related languages​​, among which the translation is made. Machine translation of literary texts have almost always poor quality.

However, in the case of technical documents translation it possible to obtain the translation of acceptable quality, which requires only minor editorial corrections if specialized machine dictionaries are in the presence and some system settings on the features of different types of texts. The more formalized style of the original document, the greater quality of the translation can be expected. While using the machine translation the best results for a text written in a technical (different descriptions and guidelines) and in official-business style can be achieved.

1. Before you start to translate, you must determine:
1) the purpose of the translation results;
2) what the source text is.

Purpose of translation plays a key role in the assessment of its quality. The same result can be considered perfect if you just need to find out what was going on in the original article, and absolutely unusable if you want the text to be published in a book or magazine. But sometimes even the "rough" translation is acceptable if it has enough information for the expert in the subject field for easily recovering the text.

On the other hand, determining what style of speech the source text belongs to, it is easy to assess its suitability for machine translation. The more allegories, metaphors in the text, and the more free style, the worse your computer manages with the translation.

Scientific, technical and educational texts are translated better then other, where the material are presentated rigorously. If special dictionaries to fill with new terms in time, you can get a fully coherent translations, requiring a minimal stylistic correlations.

Elementary and journalistic style, where are many specific patterns, but most of the words used in the direct meaning, is suitable for superficial translation. But manual correction is required to obtain literate target text. Poetry and fiction are completely unsuitable for machine translation.

2. Before the beginning of translation, it’s necessary to check the source text for any typos.

Very often the cause of an incorrect translation typos are in the original, particularly in scaned and recognized texts. In most cases words with spelling mistakes marked by the system as unknown. If a typo makes one word into another, which also exists in a foreign language, the program translates it, but the meaning of the text is incorrect. Serious problems are errors in punctuation. Improperly posed comma can significantly distort the translation of the sentence.

3. It’s necessary to find and enable thematic dictionaries on the subject of the source text.

If the subject is not exactly required, it is necessary to determine the most suitable combination of available dictionaries and always create your own. Ideally, you should have your own dictionary for each text. The productive way of work implies dividing the most spread themes into subclauses. For example, you can create "Office software", "Graphics", "web", "Internet and WWW", etc. dictionaries within the scope of computer theme.

in the framework of computer topics, you can create dictionaries.

4. It’s necessary to organize the hierarchy of dictionaries in order from the particular to the general.

If there are several enabled dictionaries in the system, the quality of translation depends on the sequence of words in which the program runs. The thematical dictionary, created specially for this text must have the highest priority. Then program must request usual thematical dictionaries and only then – general dictionary. For example, translating the text about Adobe Photoshop program, the prior dictionary is “Photoshop” (because it was created specially for this text), then – “Computer Graphics”, “Informatics” and only then – the general one.

5. The whole text should not be translated at once.

There will be several words, that dictionaries don’t comprise and lots of words, which are translated inappropriately. After enabling dictionaries and assigning other options of translation, it is recommended to translate a small piece at the beginning of the text, such as the first paragraph. Then, if there are words, translated inappropriately, the translator should save them in the prior dictionary and translate the fragment again. If the result is satisfactory, you can proceed to the next paragraph. Usually you have to translate a fourth or even a third of material for exact program adjustment. Only then you can proceed to automatic text processing.

6. It is necessary to distinguish the general and the particular.

Before puting a new word into user’s dictionary, you must estimate how many times it will be translated in the text in this way. The main point in machine translation is to distinguish between general and special lexis. Preferably, the translator must know the language and subject field of the text. The best practice is to put only words that you see systematically. Otherwise, the word throughout the text will be translated incorrectly.

Following the rules provide you with higher quality of machine translation. Though, there’s no guarantee, that the translated texts are correct and stylistically adequate. However, the first task - to understand the meaning of the text - is resolved.

Дата добавления: 2014-01-11