As massive amounts of bilingual texts become available, they have the potential to significantly improve the performance of computer - aided translation (CAT) systems. For each word in the source language, the CAT human user chooses the optimal word in the target language to convey the same meaning. The system facilitate s this process by providing relevant bilingual sentences retrieved from the translation memory. Although deep learning has made seminal contributions to machine translation, it has yet to make an impact on CAT. Previous work has applied word and sentence embeddings, trained by neural models, on example sentence retrieval from the translation memory. This project will cluster bilingual example sentences according to their context, and evaluate their effectiveness in identifying the optimal translation for the user. The proposed methods are expected to enable translators to work more autonomously and to adapt quickly and confidently to different domains and genres.