Introduction

fasttran is a simple tool using Translation Memory (TM) databases to automate human text translations. Unlike common Computer Aided Translation (CAT) tools (OmegaT, for instance) fasttran has no knobs to turn in the process of the translation. Text segmentation, TM operations et al. are all hidden under the cover; you just run the translator, get back a half-translated document, and continue the work with common text processor interface.

fasttran was designed to suit freelance translators who find evolved interface of existing CAT tools too complicated to learn and operate.

Requirements

At present, fasttran can work with Microsoft Word documents only, and requires Microsoft Windows and Microsoft Word to run. Depending on selected download option, additional software may be required.

Most complete installation file is fasttran-x.x.x-setup.exe (where x.x.x is version number, e.g. 0.1.0 in fasttran-0.1.0-setup.exe). This installer executable includes everything needed to run fasttran (except for Microsoft Windows and Microsoft Word, of course). It is highly recommended to take this download option. It's only drawback is bigger download and installation size.

fasttran-x.x.x.win32.exe is Python module installer. If you use Python for everyday work, you may use this option to spare several MiBs of disk space. You will need to have installed Python, pywin32 and wxPython.

fasttran-x.x.x.tar.gz is fasttran source package. If you take this one, you must know what you are doing.

Usage

Text translation with fasttran uses Translation Memory (TM) engine. Translation Memory is a database that keeps track of your previous translations. When you begin a work on new document, you let fasttran search your own TM for all sentences of this document and see how you translated similar sentences before. If fasttran finds a match, it replaces the sentence with found translation and marks it with colored background:

• No color (white background) - exact match. You did translate exactly same sentence before, and fasttran replaced the sentence with your previous translation.
• Grey background - no translation found. The sentence is left untranslated.
• Other colors - found similar translation. The sentence was replaced, but translation does not suit the original well. You need to compare the sentence with one in the source document and fix the translation. Background color reflects the match quality, from green - very good match (found sentence was very close to original) - through yellow and red to violet - very bad match (the sentences differ significantly).

Of course, "exactness" of the search result is meant in "computer" sense: the sentences "What a beautiful day!" and "What a terrible day!" are very similar: they differ with only one word; fasttran does not understand the meaning of words, it only knows how to split a text to sentences, and sentences to words.

Before you can use fasttran to translate documents, you need to create your Translation Memory database.

Find a pair of documents (source and translation) that you did before. In the fasttran window, activate the second notebook page: TM training. Select source document language from the pull-down list. When you do this the first time, your languages will be somewhere in the middle of the list, ordered alphabetically. But as you do the training, used languages pop up to the beginning of the list. Enter source document name in the next entry field or click the button on the right side and select the document in Select a file dialog. Fill in target language and document file name and push the Learn! button. If all goes well, after some time you will be present with a message window saying how many sentences were learned from this document pair.

Repeat learning for as many document pairs as you can find - the bigger are your TM databases, the better are your chances of getting a good translation.

When you are done with learning, you may try to actually translate something. In fasttran window, switch to the first page: Translation. Source and translation language selectors are set to the last language pair you used for TM training. Adjust if needed. Note that language selection lists on this page only allow to select language pairs for existing TM databases, you cannot select a language that you didn't teach to the program. Enter file name for the source document, or select source file from Select a file dialog. You may also specify target document name, but if you don't then it will be generated automatically from the source file name and target language tag. After some processing you will get a half-translated document, full of color marks. From here you're on your own.

After you have finished translating the document, don't forget to feed your translation to TM training. Even if translation is perfectly good (improbable!) and you did not change anything, learning the translation does useful things [1] to your TM database.

When you start using fasttran, perhaps first translations will be very poor. Please do not be disappointed. As your TM bases grow bigger, translation quality will improve.

Good luck!