Today, there is good MT or MT-supported translation, then why go to the trouble of creating a Translation Memory (TM) for a CAT tool from existing translations via alignment? The reason is that exactly the available verified translations are needed. MT can probably provide some of these translations by chance, but not reliably.
In this workshop I will present a customer case showing how MT can simplify and speed up the process of alignment, converting translation materials automatically and accurately into translation memory.
The Swiss government needs to align regularly over 5,000 legal documents (The Classified Compilation of Federal Legislation, in German “Systematische Rechtssammlung”, “SR” for short) in three languages (German, French, Italian) to create a multi-lingual and multi-directional translation memory. To achieve this they are using STAR’s fully automated machine alignment workflow.
As part of STAR’s Translation Technologies suite the interactive alignment tool assists language professionals semi-automatically in their alignment tasks. This tool already uses existing translations from translation memories to find the most probable alignment candidates based on the translation similarity. The resulting alignment proposals only need to be confirmed or adjusted by the professional user. But often – as in the use case with the Swiss Government – there are no suitable TMs available. And secondly and most importantly, interactive alignment – even if it delivers the best alignment results – is not an option for big data for reasons of time, effort and cost.
Therefore, the first question to ask was, how to get translations for comparison? The answer is that there is good machine translation for the languages in question and the alignment tool provides an interface to various MT systems. In this particular scenario, the quality of the machine translation as such is not as relevant as its similarity to the text in the target language document.
The second question to ask was, how to align the translation units automatically with a satisfying precision and certainty? For that purpose, STAR has developed a machine alignment tool. The algorithm uses the similarity score between the machine translation and alignment candidates in a very sophisticated way. Perfect matches are considered accurate alignments and they are used as reliable anchors to help calculate the alignment of the segments in between. With the help of fuzzy matches the target segments are marked as probable alignments. To sychronise source and target accordingly, segments are automatically joined, moved, deleted, or empty segments are inserted. Thus, mainly based on the similarity score entire documents can be automatically aligned at sentence-level with high precision and certainty.
Finally, to speed up the alignment of materials and provide a TM database as a result, the entire process runs as a fully-automated machine alignment and TM creation workflow on the STAR CLM platform.
With CLM, STAR has developed and established a corporate language management solution for all the customer’s needs. Its multi tenancy guaranties that all what happens on your desks stays on your desk. The principle of Lean User defined Intelligent Sequences gives user a toolbox of flexible components which fit to any requirement and every process.
Now it is expanded by CLM WebEdit which is a precious piece of browser application. It shortens the way from the project management to processing and reviewing translations. Translators and language experts can work together in real-time, online, anytime and anywhere. By using the powerful features that STAR products are known for, they benefit from many years of experience in professional language processing.
In this workshop you will experience the possibilities that online collaboration in translation and reviewing projects opens up for you.
We live in an increasingly globalized world. The exchange of information is becoming faster and faster with communication more and more multilingual. There is a growing demand for translations, but the time and resources available to all of us are limited. Therefore, optimization through machine translation is at the top of most companies‘ to-do list.
STAR focuses on collaboration and customization. Your data and sources enable STAR MT to deliver machine translations in your company-specific wording and style. In this workshop, I would like to show you briefly the possibilities of STAR MT with our translation memory tool Transit and introduce you to our web application STAR MT Translate.
STAR MT Translate provides a secure access to machine translation for everybody in your company.
The user interface opens via your browser and you have the possibility to use machine translation not only for individual words, sentences or short texts, but also entire documents. A wide range of standard file formats are available for document translation. Of course, customer-specific formats can also be integrated.
Is your customized and company internally used engine too specific for general translation requests of your employees? No problem, in STAR MT Translate you can also implement solutions from other MT providers.
You can even connect your WebTerm dictionaries to STAR MT Translate. In this way you can ensure that all users have the chance to work with your company terminology.
As administrator, you have the option to check usage statistics: Which language combinations are requested how often and with what volume of text? A regular statistical evaluation helps to identify the needs of your employees, to question the reasons for rarely used engines and to optimize your MT engines.
Please join me on a journey through the MT world at STAR!