We describe an open-source toolkit for neural machine translation (NMT). The toolkit prioritizes efficiency, modularity, and extensibility with the goal of supporting NMT research into model architectures, feature representations, and source modalities, while maintaining competitive performance and reasonable training requirements. The toolkit consists of modeling and translation support, as well as detailed pedagogical documentation about the underlying techniques.
We formulate sequence to sequence transduction as a noisy channel decoding problem and use recurrent neural networks to parameterise the source and channel models. Unlike direct models which can suffer from explaining-away effects during training, noisy channel models must produce outputs that explain their inputs, and their component models can be trained with not only paired training samples but also unpaired samples from the marginal output distribution. Using a latent variable to control how much of the conditioning sequence the channel model needs to read in order to generate a subsequent symbol, we obtain a tractable and effective beam search decoder. Experimental results on abstractive sentence summarisation, morphological inflection, and machine translation show that noisy channel models outperform direct models, and that they significantly benefit from increased amounts of unpaired output data that direct models cannot easily use.
Several mechanisms to focus attention of a neural network on selected parts of its input or memory have been used successfully in deep learning models in recent years. Attention has improved image classification, image captioning, speech recognition, generative models, and learning algorithmic tasks, but it had probably the largest impact on neural machine translation. Recently, similar improvements have been obtained using alternative mechanisms that do not focus on a single part of a memory but operate on all of it in parallel, in a uniform way. Such mechanism, which we call active memory, improved over attention in algorithmic tasks, image processing, and in generative modelling. So far, however, active memory has not improved over attention for most natural language processing tasks, in particular for machine translation. We analyze this shortcoming in this paper and propose an extended model of active memory that matches existing attention models on neural machine translation and generalizes better to longer sentences. We investigate this model and explain why previous active memory models did not succeed. Finally, we discuss when active memory brings most benefits and where attention can be a better choice.
Despite recent advances, memory-augmented deep neural networks are still limited when it comes to life-long and one-shot learning, especially in remembering rare events. We present a large-scale life-long memory module for use in deep learning. The module exploits fast nearest-neighbor algorithms for efficiency and thus scales to large memory sizes. Except for the nearest-neighbor query, the module is fully differentiable and trained end-to-end with no extra supervision. It operates in a life-long manner, i.e., without the need to reset it during training. Our memory module can be easily added to any part of a supervised neural network. To show its versatility we add it to a number of networks, from simple convolutional ones tested on image classification to deep sequence-to-sequence and recurrent-convolutional models. In all cases, the enhanced network gains the ability to remember and do life-long one-shot learning. Our module remembers training examples shown many thousands of steps in the past and it can successfully generalize from them. We set new state-of-the-art for one-shot learning on the Omniglot dataset and demonstrate, for the first time, life-long one-shot learning in recurrent neural networks on a large-scale machine translation task.
In neural machine translation (NMT), generation of a target word depends on both source and target contexts. We find that source contexts have a direct impact on the adequacy of a translation while target contexts affect the fluency. Intuitively, generation of a content word should rely more on the source context and generation of a functional word should rely more on the target context. Due to the lack of effective control over the influence from source and target contexts, conventional NMT tends to yield fluent but inadequate translations. To address this problem, we propose context gates which dynamically control the ratios at which source and target contexts contribute to the generation of target words. In this way, we can enhance both the adequacy and fluency of NMT with more careful control of the information flow from contexts. Experiments show that our approach significantly improves upon a standard attention-based NMT system by +2.3 BLEU points.
We present in this paper our work on comparison between Statistical Machine Translation (SMT) and Rule-based machine translation for translation from Marathi to Hindi. Rule Based systems although robust take lots of time to build. On the other hand statistical machine translation systems are easier to create, maintain and improve upon. We describe the development of a basic Marathi-Hindi SMT system and evaluate its performance. Through a detailed error analysis, we, point out the relative strengths and weaknesses of both systems. Effectively, we shall see that even with a small amount of training corpus a statistical machine translation system has many advantages for high quality domain specific machine translation over that of a rule-based counterpart.
By capturing statistical patterns in large corpora, machine learning has enabled significant advances in natural language processing, including in machine translation, question answering, and sentiment analysis. However, for agents to intelligently interact with humans, simply capturing the statistical patterns is insufficient. In this paper we investigate if, and how, grounded compositional language can emerge as a means to achieve goals in multi-agent populations. Towards this end, we propose a multi-agent learning environment and learning methods that bring about emergence of a basic compositional language. This language is represented as streams of abstract discrete symbols uttered by agents over time, but nonetheless has a coherent structure that possesses a defined vocabulary and syntax. We also observe emergence of non-verbal communication such as pointing and guiding when language communication is unavailable.
We present Nematus, a toolkit for Neural Machine Translation. The toolkit prioritizes high translation accuracy, usability, and extensibility. Nematus has been used to build top-performing submissions to shared translation tasks at WMT and IWSLT, and has been used to train systems for production environments.
We present a novel neural network for processing sequences. The ByteNet is a one-dimensional convolutional neural network that is composed of two parts, one to encode the source sequence and the other to decode the target sequence. The two network parts are connected by stacking the decoder on top of the encoder and preserving the temporal resolution of the sequences. To address the differing lengths of the source and the target, we introduce an efficient mechanism by which the decoder is dynamically unfolded over the representation of the encoder. The ByteNet uses dilation in the convolutional layers to increase its receptive field. The resulting network has two core properties: it runs in time that is linear in the length of the sequences and it sidesteps the need for excessive memorization. The ByteNet decoder attains state-of-the-art performance on character-level language modelling and outperforms the previous best results obtained with recurrent networks. The ByteNet also achieves state-of-the-art performance on character-to-character machine translation on the English-to-German WMT translation task, surpassing comparable neural translation models that are based on recurrent networks with attentional pooling and run in quadratic time. We find that the latent alignment structure contained in the representations reflects the expected alignment between the tokens.
Neural Machine Translation (NMT) has shown remarkable progress over the past few years with production systems now being deployed to end-users. One major drawback of current architectures is that they are expensive to train, typically requiring days to weeks of GPU time to converge. This makes exhaustive hyperparameter search, as is commonly done with other neural network architectures, prohibitively expensive. In this work, we present the first large-scale analysis of NMT architecture hyperparameters. We report empirical results and variance numbers for several hundred experimental runs, corresponding to over 250,000 GPU hours on the standard WMT English to German translation task. Our experiments lead to novel insights and practical advice for building and extending NMT architectures. As part of this contribution, we release an open-source NMT framework that enables researchers to easily experiment with novel techniques and reproduce state of the art results.
In state-of-the-art Neural Machine Translation, an attention mechanism is used during decoding to enhance the translation. At every step, the decoder uses this mechanism to focus on different parts of the source sentence to gather the most useful information before outputting its target word. Recently, the effectiveness of the attention mechanism has also been explored for multimodal tasks, where it becomes possible to focus both on sentence parts and image regions. Approaches to pool two modalities usually include element-wise product, sum or concatenation. In this paper, we evaluate the more advanced Multimodal Compact Bilinear pooling method, which takes the outer product of two vectors to combine the attention features for the two modalities. This has been previously investigated for visual question answering. We try out this approach for multimodal image caption translation and show improvements compared to basic combination methods.
We report the results of our classification-based machine translation model, built upon the framework of a recurrent neural network using gated recurrent units. Unlike other RNN models that attempt to maximize the overall conditional log probability of sentences against sentences, our model focuses a classification approach of estimating the conditional probability of the next word given the input sequence. This simpler approach using GRUs was hoped to be comparable with more complicated RNN models, but achievements in this implementation were modest and there remains a lot of room for improving this classification approach.
Machine translation (MT) was developed as one of the hottest research topics in the natural language processing (NLP) literature. One important issue in MT is that how to evaluate the MT system reasonably and tell us whether the translation system makes an improvement or not. The traditional manual judgment methods are expensive, time-consuming, unrepeatable, and sometimes with low agreement. On the other hand, the popular automatic MT evaluation methods have some weaknesses. Firstly, they tend to perform well on the language pairs with English as the target language, but weak when English is used as source. Secondly, some methods rely on many additional linguistic features to achieve good performance, which makes the metric unable to replicate and apply to other language pairs easily. Thirdly, some popular metrics utilize incomprehensive factors, which result in low performance on some practical tasks. In this thesis, to address the existing problems, we design novel MT evaluation methods and investigate their performances on different languages. Firstly, we design augmented factors to yield highly accurate evaluation. Secondly, we design a tunable evaluation model where weighting of factors can be optimized according to the characteristics of languages. Thirdly, in the enhanced version of our methods, we design concise linguistic feature using part-of-speech (POS) to show that our methods can yield even higher performance when using some external linguistic resources. Finally, we introduce the practical performance of our metrics in the ACL-WMT workshop shared tasks, which show that the proposed methods are robust across different languages. In addition, we also present some novel work on quality estimation of MT without using reference translations including the usage of probability models of Na\"ive Bayes (NB), support vector machine (SVM) classification algorithms, and CRFs.
I started this work with the hope of generating a text synthesizer (like a musical synthesizer) that can imitate certain linguistic styles. Most of the report focuses on text simplification using statistical machine translation (SMT) techniques. I applied MOSES to a parallel corpus of the Bible (King James Version and Easy-to-Read Version) and that of Wikipedia articles (normal and simplified). I report the importance of the three main components of SMT---phrase translation, language model, and recording---by changing their weights and comparing the resulting quality of simplified text in terms of METEOR and BLEU. Toward the end of the report will be presented some examples of text "synthesized" into the King James style.
One of the most important problems in machine translation (MT) evaluation is to evaluate the similarity between translation hypotheses with different surface forms from the reference, especially at the segment level. We propose to use word embeddings to perform word alignment for segment-level MT evaluation. We performed experiments with three types of alignment methods using word embeddings. We evaluated our proposed methods with various translation datasets. Experimental results show that our proposed methods outperform previous word embeddings-based methods.
In this paper, we explore a simple solution to "Multi-Source Neural Machine Translation" (MSNMT) which only relies on preprocessing a N-way multilingual corpus without modifying the Neural Machine Translation (NMT) architecture or training procedure. We simply concatenate the source sentences to form a single long multi-source input sentence while keeping the target side sentence as it is and train an NMT system using this preprocessed corpus. We evaluate our method in resource poor as well as resource rich settings and show its effectiveness (up to 4 BLEU using 2 source languages and up to 6 BLEU using 5 source languages). We also compare against existing methods for MSNMT and show that our solution gives competitive results despite its simplicity. We also provide some insights on how the NMT system leverages multilingual information in such a scenario by visualizing attention.
Artificial neural networks have been successfully applied to a variety of machine learning tasks, including image recognition, semantic segmentation, and machine translation. However, few studies fully investigated ensembles of artificial neural networks. In this work, we investigated multiple widely used ensemble methods, including unweighted averaging, majority voting, the Bayes Optimal Classifier, and the (discrete) Super Learner, for image recognition tasks, with deep neural networks as candidate algorithms. We designed several experiments, with the candidate algorithms being the same network structure with different model checkpoints within a single training process, networks with same structure but trained multiple times stochastically, and networks with different structure. In addition, we further studied the over-confidence phenomenon of the neural networks, as well as its impact on the ensemble methods. Across all of our experiments, the Super Learner achieved best performance among all the ensemble methods in this study.
Neural machine translation (NMT), a new approach to machine translation, has achieved promising results comparable to those of traditional approaches such as statistical machine translation (SMT). Despite its recent success, NMT cannot handle a larger vocabulary because training complexity and decoding complexity proportionally increase with the number of target words. This problem becomes even more serious when translating patent documents, which contain many technical terms that are observed infrequently. In NMTs, words that are out of vocabulary are represented by a single unknown token. In this paper, we propose a method that enables NMT to translate patent sentences comprising a large vocabulary of technical terms. We train an NMT system on bilingual data wherein technical terms are replaced with technical term tokens; this allows it to translate most of the source sentences except technical terms. Further, we use it as a decoder to translate source sentences with technical term tokens and replace the tokens with technical term translations using SMT. We also use it to rerank the 1,000-best SMT translations on the basis of the average of the SMT score and that of the NMT rescoring of the translated sentences with technical term tokens. Our experiments on Japanese-Chinese patent sentences show that the proposed NMT system achieves a substantial improvement of up to 3.1 BLEU points and 2.3 RIBES points over traditional SMT systems and an improvement of approximately 0.6 BLEU points and 0.8 RIBES points over an equivalent NMT system without our proposed technique.
This paper investigates the robustness of NLP against perturbed word forms. While neural approaches can achieve (almost) human-like accuracy for certain tasks and conditions, they often are sensitive to small changes in the input such as non-canonical input (e.g., typos). Yet both stability and robustness are desired properties in applications involving user-generated content, and the more as humans easily cope with such noisy or adversary conditions. In this paper, we study the impact of noisy input. We consider different noise distributions (one type of noise, combination of noise types) and mismatched noise distributions for training and testing. Moreover, we empirically evaluate the robustness of different models (convolutional neural networks, recurrent neural networks, non-neural models), different basic units (characters, byte pair encoding units), and different NLP tasks (morphological tagging, machine translation).
Chinese discourse coherence modeling remains a challenge taskin Natural Language Processing field.Existing approaches mostlyfocus on the need for feature engineering, whichadoptthe sophisticated features to capture the logic or syntactic or semantic relationships acrosssentences within a text.In this paper, we present an entity-drivenrecursive deep modelfor the Chinese discourse coherence evaluation based on current English discourse coherenceneural network model. Specifically, to overcome the shortage of identifying the entity(nouns) overlap across sentences in the currentmodel, Our combined modelsuccessfully investigatesthe entities information into the recursive neural network freamework.Evaluation results on both sentence ordering and machine translation coherence rating task show the effectiveness of the proposed model, which significantly outperforms the existing strong baseline.