2016/05/16 by Marcin Junczys-Dowmunt, Junczys-Dowmunt, Marcin, Tomasz Dwojak +3 · 1 citation
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1605.04809
openalex publication_date 2016/05/16 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
This paper describes the AMU-UEDIN submissions to the WMT 2016 shared task on\nnews translation. We explore methods of decode-time integration of\nattention-based neural translation models with phrase-based statistical machine\ntranslation. Efficient batch-algorithms for GPU-querying are proposed and\nimplemented. For English-Russian, our system stays behind the state-of-the-art\npure neural models in terms of BLEU. Among restricted systems, manual\nevaluation places it in the first cluster tied with the pure neural model. For\nthe Russian-English task, our submission achieves the top BLEU result,\noutperforming the best pure neural system by 1.1 BLEU points and our own\nphrase-based baseline by 1.6 BLEU. After manual evaluation, this system is the\nbest restricted system in its own cluster. In follow-up experiments we improve\nresults by additional 0.8 BLEU.\n