Using Manual and Parallel Aligned Corpora for Machine Translation Services within an On-line Content Management System

Web content management systems (WCMSs) are a popular instrument for gathering, navigating and assessing information in environments such as Digital Libraries or e-Learning. Such environments are characterized not only through a critical amount of documents, but also by their domain heterogeneity, relative to format, domain or date of production, and their multilingual character. Methods from Information and Language Technology are the “plug-ins” necessary to any WCMS in order to ensure a proper functionality, given the features mentioned above. Among these “plug-ins”, machine translation (MT) is a key component, which enables translation of meta-data and content either for the user or for other components of the WCMS (i.e. crosslingual retrieval component). However, the MT task is extremely challenging and lacks frequently the availability of adequate training data. In this paper we will present a WCMS including machine translation, explain the related MT challenges, and discuss the employment of corpora as training material, which are manually and automatically parallel aligned.