Do What I Mean: Issues in Cross-lingual Collaboration

In most virtual worlds and virtual communities, language is at the heart of communication. When we extend these communities to the international stage, we are faced with challenges in interaction. We will not examine the infrastructure supports beneath multi- and cross-lingual communication. Instead, we look at the next level of enabling communication—that of automated machine translation (MT) of the utterances. MT is a complex and, it will be argued here, an endogenous problem. Even though the MT process is endogenous, we are not prevented from creating useful interfaces and supports for cross language interaction in virtual worlds and simulations. This paper starts with a basic description of machine translation, the problems of MT development and why MT can be categorized as endogenous. Afterwards, we describe the classes of problems that MT must solve which are specific to virtual words. Then, we look deeply into the communicative system and start to describe a framework to handle one particularly pesky problem of MT: the process of building language-independent representations to enable effective translation.