Trainable Greedy Decoding for Neural Machine Translation

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1968–1978 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Trainable Greedy Decoding for Neural Machine Translation Jiatao Gu †, Kyunghyun Cho ‡ and Victor O.K. Li † †The University of Hong Kong ‡New York University †{jiataogu, vli}@eee.hku.hk ‡kyunghyun.cho@nyu.edu Abstract Recent research in neural machine trans- lation has largely focused on two aspects; neural network architectures and end-to- end learning algorithms. The problem of decoding, however, has received relatively little attention from the research commu- nity. In this paper, we solely focus on the problem of decoding given a trained neu- ral machine translation model. Instead of trying to build a new decoding algorithm for any specific decoding objective, we pro- pose the idea of trainable decoding algo- rithm in which we train a decoding algo- rithm to find a translation that maximizes an arbitrary decoding objective. More specifically, we design an actor that ob- serves and manipulates the hidden state of the neural machine translation decoder and propose to train it using a variant of deterministic policy gradient. We exten- sively evaluate the proposed algorithm us- ing four language pairs and two decoding objectives, and show that we can indeed train a trainable greedy decoder that gener- ates a better translation (in terms of a target decoding objective) with minimal compu- tational overhead. 1 Introduction Neural machine translation has recently become a method of choice in machine translation research. Besides its success in traditional settings of ma- chine translation, that is one-to-one translation be- tween two languages, (Sennrich et al., 2016; Chung et al., 2016), neural machine translation has ven- tured into more sophisticated settings of machine translation. For instance, neural machine transla- tion has successfully proven itself to be capable of handling subword-level representation of sentences (Lee et al., 2016; Luong and Manning, 2016; Sen- nrich et al., 2015; Costa-Jussa and Fonollosa, 2016; Ling et al., 2015). Furthermore, several research groups have shown its potential in seamlessly han- dling multiple languages (Dong et al., 2015; Luong et al., 2015a; Firat et al., 2016a,b; Lee et al., 2016; Ha et al., 2016; Viégas et al., 2016). A typical scenario of neural machine transla- tion starts with training a model to maximize its log-likelihood. That is, we often train a model to maximize the conditional probability of a reference translation given a source sentence over a large par- allel corpus. Once the model is trained in this way, it defines the conditional distribution over all pos- sible translations given a source sentence, and the task of translation becomes equivalent to finding a translation to which the model assigns the highest conditional probability. Since it is computationally intractable to do so exactly, it is a usual practice to resort to approximate search/decoding algorithms such as greedy decoding or beam search. In this scenario, we have identified two points where im- provements could be made. They are (1) training (including the selection of a model architecture) and (2) …

Content truncated to 500 words