University of Rochester WMT 2017 NMT System Submission

Proceedings of the Conference on Machine Translation (WMT), Volume 2: Shared Task Papers, pages 310–314 Copenhagen, Denmark, September 711, 2017. c⃝2017 Association for Computational Linguistics University of Rochester WMT 2017 NMT System Submission Chester Holtz, Chuyang Ke, and Daniel Gildea University of Rochester choltz2@u.rochester.edu Abstract We describe the neural machine trans- lation system submitted by the Univer- sity of Rochester to the Chinese-English language pair for the WMT 2017 news translation task. We applied unsuper- vised word and subword segmentation techniques and deep learning in order to address (i) the word segmentation prob- lem caused by the lack of delimiters be- tween words and phrases in Chinese and (ii) the morphological and syntactic differ- ences between Chinese and English. We integrated promising recent developments in NMT, including back-translations, lan- guage model reranking, subword splitting and minimum risk tuning. 1 Introduction This paper presents the machine translation (MT) systems submitted by University of Rochester to the WMT 2017 news translation task. We partic- ipated in the Chinese-to-English and Latvian-to- English news translation tasks, but will focus on describing the system submitted for the Chinese- to-English task. Chinese-to-English is a particularly challenging language pair for corpus-based MT systems due to the task of finding an optimal word segmenta- tion for Chinese sentences as well as other linguis- tic differences between Chinese and English sen- tences. For example the fact that there may exist multiple possible meanings for characters depend- ing on their context and that individual characters can be joined together to build compound words exacerbate the aforementioned segmentation prob- lem. Additionally, translation performance is also affected by the frequent dropping of subjects and infrequent use of function words in Chinese sen- tences. We used both word-level and morphological feature-based representations of Chinese to deal with data sparsity and reduce the size of the Chi- nese vocabulary. We experimented with both sub- phrase-based and character-based systems. Both RNN-based and 5-gram language models were trained with data extracted from the English news corpora provided and are used to rerank hypothe- ses proposed by the decoder. The paper is organized as follows: in Section 2 we introduce our system and preprocessing meth- ods for the Chinese language. Our main learning framework training settings are explained in Sec- tion 3. Our NMT, SMT, and submission results are presented in Section 4. The paper ends with some concluding remarks. 2 System Description In this section we briefly introduce our prepro- cessing methods and the general encoder-decoder framework with attention (Sutskever et al., 2014; Cho et al., 2014; Bahdanau et al., 2014) used in our system. We closely followed the neural ma- chine translation model proposed by Chorowski et al. (2015). A neural machine translation model (Kalch- brenner and Blunsom, 2013; Cho et al., 2014; Sutskever et al., 2014) aims at building an end- to-end neural network framework, which takes as input a source sentence X = (x1, ..., xTX) with length of TX, and outputs its translation Y = (y1, ..., yTY ) with length of …

Content truncated to 500 words