myCat

About Computer-Assisted Translation (CAT)

In most translation contexts, be it international organizations or private companies, some parts of the documents to be translated are actually repetitions of other documents which were previously translated. Besides, each organization has its own terminology and style which must be respected for the sake of precision and consistency.

Thus translators need tools to detect whether a given part of a text was already translated or not, and to let them know how a specific concept should be translated in a specific context for a specific customer. This is what Computer-Assisted Translation (CAT) is about: providing software tools to translators to help them improve the accuracy and consistency of their work and to support their productivity.

CAT tools generally include bi-text alignment tools (see below), translation memories and terminological management tools, as well as supporting tools such as terminology extractors, word counting tools, format converters, and even translation management solutions, interfaces with electronic document management systems, etc.

myCat: A Concordancer composed of a Text Aligner and a Quote Detector

The Text Aligner :

The Quote Detector :

myCat was developed by Simple Shift at the request of several international organizations who were looking for an extremely simple tool to re-use their previously-translated documents as references for the next translations. It is currently in beta-testing phase and includes the following modules:

• A bi-text aligner, which allows finding terms and expressions in full-text documents and aligning automatically the source and target versions of the reference documents. Thus it provides translators with the term in context, in both the source and target reference documents.
• A quote detector, which allows detecting all the parts of a text which were previously translated, and shows the corresponding documents in source and target versions with an automatic alignment.

The core value of myCat is its automatic alignment of the source and target versions: It avoids translators looking manually into the source document, then finding and opening the target language equivalent and browsing it until they find the translation. It saves a lot of research time!

What is unique about myCat is its alignment technology: It is built on the statistical machine translation tool called Moses (see the myMT page), which we use to build translation maps for every document pair in the reference corpus in all the customer's working languages. Translation maps allow the system to know which source sentence contains the required term and which sentence is the most probable translation in the target document. This break-through technology is not perfect since it relies on probabilistic models, but it is in average much more efficient than simpler alignment approaches such as working on document proportions, counting full-stop marks, looking at paragraph numbering and figures, etc.

Version 1.0 of myCat is available since November 2011.

More modules should be added to myCat over time, including a terminological management tool and a translation memory tool. Interfaces should also be developed with myMT in order to integrate the MT and CAT tools in a single platform, and with open-source and other CMS tools so as to allow for adequate document management.