Translation, the core business of AzTech Solutions, is the act of rephrasing text in another language to convey the same meaning, style and atmosphere as the source text. The translation process requires an understanding of the target language at native speaker level and nearly the same for the source language. Every sentence has to be read, understood and rephrased in the target language.

The computer era has radically changed the translation process. It has brought about two related but different types of software: machine translation (MT) software and computer-aided translation software (CAT tools). Huge text corpora have improved the quality of machine translation, but the MT software currently available still cannot provide the same standard of quality as a qualified human translator. [+]Machine translation was first based on the substitution of a single word by another, using a statistics-based method to identify the most common translation for that word. Nowadays the machine translation software also takes the adjacent words into consideration and even attempts to translate as many words as possible as a single replaceable phrase before cutting it into parts. The results are best when the text corpora consist of relevant texts for a specific field, not just any text. Compiling such corpora is a time-consuming procedure that is only useful when it is to be used for huge quantities of text, i.e. for very large projects of at least half a million words. Sometimes a human reviewer corrects the first batches of the translation and this text is added to the corpora to increase the quality of the next batches. Machine translation has proven useful when applied to specific professions, for example weather reports. The quality of the output of machine translation software is improving, but the technology still has a long way to go.

CAT tools have increased the quantity of words that can be processed by a translator when using a large appropriate memory (database of previously translated segments) to translate similar texts. This is especially useful when translating a new version of technical manuals or legal texts, but is much less so for literature and other creative texts – although the concordance feature may be useful when trying to find a previous translation for a specific term. There are many CAT tools available and each has its own group of enthusiastic fans. [+]The term CAT tool is often used in the translation industry to refer to the various translation memory software applications, such as Déjà Vu, memoQ, SDL Passolo, SDL Trados, SDL WorldServer and WordFast. This is because modern translation memory software often offers one or more of the following additional tools: language search engine software, terminology management software, alignment software and interactive machine translation. Some CAT tools include all of these options, while others do not.

A CAT tool is used at the beginning of the translation process to establish the amount of work involved. The software analyses the text to be translated against the relevant memory and determines the number of words to be translated. To do this, the text is broken into segments. Each segment is compared to the segments in the memory. [+]A segment is text separated by specific symbols and/or punctuation marks (e.g. period, semicolon, colon, etc.); some translation software packages allow the user to define which punctuation marks should be used to mark the end of a segment.

The similarity between a new segment and a memory segment is indicated as a percentage, ranging from 100% for identical segments (or full matches) to 0% for totally new segments. Segments which are moderately or highly similar to a memory segment are called fuzzy matches.[+]Segments are usually listed in match percentage groups, for example: 0-74% for new segments, 75-99% for fuzzy matches, 100% for full matches.

The tool also indicates whether the new segments include internal repetitions. Most tools distinguish two different types of full matches: the regular full match, which simply means that the current segment is exactly the same, and the in-context exact match (ICE match) which is sometimes referred to as a 101% match.[+]An ICE match exactly matches a memory segment, and also occurs in exactly the same context, which usually means that the surrounding segments are identical to those in the previous text as well.
All segments which are below a certain threshold for a match to previous work in the memory are considered to be new segments.

The analysis of the work is used to generate the quotation to be sent to the customer. The quotation usually only indicates the weighted amount of words for the project. The weighted word quantity is based on the graduated rate ladder used for the translation. The rate ladder specifies the percentage of the word rate to be applied to the fuzzy and full matches in the translation. This makes quotations easier to read. Sometimes the total number of words is also provided for information purposes.