With more than 300,000 employees in 75 countries, our client, Veolia Environnement, decided to introduce Machine Translation (MT) in order to reduce the costs of translating internal documents. Specifically, the company had initiated a project to deploy SAP globally, with simultaneous roll‐out in French and in English. Tight deadlines necessitated nearly instantaneous translation of the SAP procedures manuals into English as they were being written in French.
The client had already chosen Systran’s Machine Translation engine, so before the launch of the SAP project, LexWorks’s task was to “train” the MT engine on SAP terminology. A terminologist was sent to the client’s premises to work alongside the client’s SAP team to build domain‐ and company‐specific dictionaries. Building the dictionaries was the first step in training the MT engine for Veolia. The next step was to test the performance of the engine by translating typical source documents followed by immediate post‐editing of the resulting translations. By feeding the corrections back into the system, LexWorks was able to gradually train the MT engine to produce better and better translations. This process was accompanied by a strong framework of progress reports and process documentation by LexWorks’s MT project manager
For the SAP roll‐out, our client decided to combine Machine Translation with human post‐editing to ensure the quality they needed within their extremely tight time‐frame. Two LexWorks terminologists worked alongside the SAP staff, each with the task of post‐editing one SAP manual per day. In the first phase of the project, two terminologists were able to translate and perform light post‐editing of 394,000 words in one month. A project of that scope, which would under normal circumstances have required 10 translators and one editor, was completed on time, with only two post‐editors, thus saving our client over $30K and 103 person days.