A wide range of applications play an important role in the daily work of the modern human translator. However, the computational tools designed to aid in the process of translation only benefit translation from or to a small minority of the 7,000 languages of the world, those that we may call "privileged languages". As for those translators who work with the remaining languages, the marginalized languages in the digital world, they cannot benefit from the tools that are speeding up the production of translation in the privileged languages. We may ask whether it is possible to bridge the gap between what is available for these languages and for the marginalized ones. This paper proposes a framework for computer-assisted translation into marginalized languages and its implementation in a web application for Spanish-Guarani translation. The proposed system is based on a new theory for phrase-level translation in contexts where adequate bilingual corpora are not available: Translation by Generalized Segments (referred to as Minimal Dependency Translation in previous work).