Conventional retrieval-augmented neural machine translation (RANMT) systems leverage bilingual corpora, e.g., translation memories (TMs). Yet, in many settings, monolingual corpora in the target language are often available. This work explores ways to take advantage of such resources by directly retrieving relevant target language segments, based on a source-side query. For this, we design improved cross-lingual retrieval systems, trained with both sentence level and word-level matching objectives. In our experiments with two RANMT architectures, we assess of such cross-lingual objectives in a controlled setting, reaching performances that match those of standard TM-based models. We also showcase our method on real-world settings, using much larger monolingual corpora, and observe strong improvements over both the baseline setting, and general-purpose cross-lingual retrievers.

Fault tolerance is an important issue for the design of interconnection networks. In this paper, a new fault tolerant routing algorithm is presented and is applied in mesh networks employing wormhole switching. Due to its low routing restrictions, the presented routing algorithm is so highly adaptive that it is connected and deadlock free in spite of the various fault regions in mesh networks. Due to the minimal virtual channels it uses, the presented routing algorithm only employs as few buffers as possible and is suitable for fault-tolerant interconnection networks with low cost. Since it chooses the path around fault regions according to the local fault information, the presented routing algorithm takes routing decisions quickly and is applicable in interconnection networks. Moreover, a simulation is conducted for the proposed routing algorithm and the results show that the algorithm exhibits a graceful degradation in performance.