This book addresses the use of comparable corpora for computer-assisted translation. The first part of the book contains a state-of-the art of the techniques used for the extraction of bilingual lexicon from comparable corpora as well as a user-centered evaluation of these lexicons. The second part of the book details the experimentation of a new method for extracting lexicons from comparable corpora. The method is based on the compositionality principle. We also detail the experimentation of learning-to-rank algorithms for ranking the generated translations. Although the usefulness of comparable corpora had been demonstrated in areas like machine translation or cross-lingual information retrieval, it had never been assessed for computer-assisted translation. Moreover, there is no book offering a state-of-the art of the techniques used in comparable corpora processing. The research, though addressing theoretical aspects, also has an applicative orientation since it was partly conducted in an industrial context.