Extração de vocabulário multilíngue a partir de documentação de software

Abstract

Real-time machine translation tools and services have been investigated as an alternative approach to the utilization of a common language (lingua franca) during distributed meetings involving teams with differet native languages. However, as presented by different research works, this kind of technologies presents a set of problems that difficults the communication. Among the solution proposed in the literature, the construction of domain specific vocabularies are highlited. This work propose a multilingual vocabulary extraction process for multilingual dicionary entries extraction from software user guides. The process here proposed follows a well stablished set of steps presenting as the main difference the way in wich the domain vocabulary is identified: through the utilization of terminology extraction softwares. A manual evaluation of the dictionaries generated by the process has shown a precision of 81% for simple world translation and 39% for multiword expressions. This values are consistent with the related work.