In this paper we present new research in translation assistance. Wedescribe a system capable of translating native language (L1)fragments to foreign language (L2) fragments in an L2context. Practical applications of this research can be framed inthe context of second language learning. The type of translationassistance system under investigation here encourages languagelearners to write in their target language while allowing them tofall back to their native language in case the correct word orexpression is not known. These code switches are subsequentlytranslated to L2 given the L2 context. We study thefeasibility of exploiting cross-lingual context to obtainhigh-quality translation suggestions that improve over statisticallanguage modelling and word-sense disambiguation baselines. Aclassification-based approach is presented that is indeed found toimprove significantly over these baselines by making use of acontextual window spanning a small number of neighbouring words.