PyTorchKR Papers

최신 논문

총 5,206편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2009.07489

Graph-to-Sequence Neural Machine Translation

Neural machine translation (NMT) usually works in a seq2seq learning way by viewing either source or target sentence as a linear sequence of words, which can be regarded as a special case of graph, taking words in the sequence as nodes and relationships between words as edges. In the light of the current NMT models more or less capture graph information among the sequence in a latent way, we present a graph-to-sequence model facilitating explicit graph information capturing. In detail, we propose a graph-based SAN-based NMT model called Graph-Transformer by capturing information of subgraphs of different orders in every layers. Subgraphs are put into different groups according to their orders, and every group of subgraphs respectively reflect different levels of dependency between words. For fusing subgraph representations, we empirically explore three methods which weight different groups of subgraphs of different orders. Results of experiments on WMT14 English-German and IWSLT14 German-English show that our method can effectively boost the Transformer with an improvement of 1.1 BLEU points on WMT14 English-German dataset and 1.0 BLEU points on IWSLT14 German-English dataset.

arXiv:2009.07453

Extremely Low Bit Transformer Quantization for On-Device Neural Machine Translation

The deployment of widely used Transformer architecture is challenging because of heavy computation load and memory overhead during inference, especially when the target device is limited in computational resources such as mobile or edge devices. Quantization is an effective technique to address such challenges. Our analysis shows that for a given number of quantization bits, each block of Transformer contributes to translation quality and inference computations in different manners. Moreover, even inside an embedding block, each word presents vastly different contributions. Correspondingly, we propose a mixed precision quantization strategy to represent Transformer weights by an extremely low number of bits (e.g., under 3 bits). For example, for each word in an embedding block, we assign different quantization bits based on statistical property. Our quantized Transformer model achieves 11.8$\times$ smaller model size than the baseline model, with less than -0.5 BLEU. We achieve 8.3$\times$ reduction in run-time memory footprints and 3.5$\times$ speed up (Galaxy N10+) such that our proposed compression strategy enables efficient implementation for on-device NMT.

arXiv:2009.07310공식 코드 · Papers with Code

Simultaneous Machine Translation with Visual Context

Simultaneous machine translation (SiMT) aims to translate a continuous input text stream into another language with the lowest latency and highest quality possible. The translation thus has to start with an incomplete source text, which is read progressively, creating the need for anticipation. In this paper, we seek to understand whether the addition of visual information can compensate for the missing source context. To this end, we analyse the impact of different multimodal approaches and visual features on state-of-the-art SiMT frameworks. Our results show that visual context is helpful and that visually-grounded models based on explicit object region information are much better than commonly used global features, reaching up to 3 BLEU points improvement under low latency scenarios. Our qualitative analysis illustrates cases where only the multimodal systems are able to translate correctly from English into gender-marked languages, as well as deal with differences in word order, such as adjective-noun placement between English and French.

arXiv:2009.07839공식 코드 · Papers with Code

Text Generation by Learning from Demonstrations

Current approaches to text generation largely rely on autoregressive models and maximum likelihood estimation. This paradigm leads to (i) diverse but low-quality samples due to mismatched learning objective and evaluation metric (likelihood vs. quality) and (ii) exposure bias due to mismatched history distributions (gold vs. model-generated). To alleviate these problems, we frame text generation as an offline reinforcement learning (RL) problem with expert demonstrations (i.e., the reference), where the goal is to maximize quality given model-generated histories. We propose GOLD (generation by off-policy learning from demonstrations): an easy-to-optimize algorithm that learns from the demonstrations by importance weighting. Intuitively, GOLD upweights confident tokens and downweights unconfident ones in the reference during training, avoiding optimization issues faced by prior RL approaches that rely on online data collection. According to both automatic and human evaluation, models trained by GOLD outperform those trained by MLE and policy gradient on summarization, question generation, and machine translation. Further, our models are less sensitive to decoding algorithms and alleviate exposure bias.

arXiv:2009.07715

Knowledge Graphs for Multilingual Language Translation and Generation

The Natural Language Processing (NLP) community has recently seen outstanding progress, catalysed by the release of different Neural Network (NN) architectures. Neural-based approaches have proven effective by significantly increasing the output quality of a large number of automated solutions for NLP tasks (Belinkov and Glass, 2019). Despite these notable advancements, dealing with entities still poses a difficult challenge as they are rarely seen in training data. Entities can be classified into two groups, i.e., proper nouns and common nouns. Proper nouns are also known as Named Entities (NE) and correspond to the name of people, organizations, or locations, e.g., John, WHO, or Canada. Common nouns describe classes of objects, e.g., spoon or cancer. Both types of entities can be found in a Knowledge Graph (KG). Recent work has successfully exploited the contribution of KGs in NLP tasks, such as Natural Language Inference (NLI) (KM et al.,2018) and Question Answering (QA) (Sorokin and Gurevych, 2018). Only a few works had exploited the benefits of KGs in Neural Machine Translation (NMT) when the work presented herein began. Additionally, few works had studied the contribution of KGs to Natural Language Generation (NLG) tasks. Moreover, the multilinguality also remained an open research area in these respective tasks (Young et al., 2018). In this thesis, we focus on the use of KGs for machine translation and the generation of texts to deal with the problems caused by entities and consequently enhance the quality of automatically generated texts.

arXiv:2009.07610공식 코드 · Papers with Code

Reusing a Pretrained Language Model on Languages with Limited Corpora for Unsupervised NMT

Using a language model (LM) pretrained on two languages with large monolingual data in order to initialize an unsupervised neural machine translation (UNMT) system yields state-of-the-art results. When limited data is available for one language, however, this method leads to poor translations. We present an effective approach that reuses an LM that is pretrained only on the high-resource language. The monolingual LM is fine-tuned on both languages and is then used to initialize a UNMT model. To reuse the pretrained LM, we have to modify its predefined vocabulary, to account for the new language. We therefore propose a novel vocabulary extension method. Our approach, RE-LM, outperforms a competitive cross-lingual pretraining model (XLM) in English-Macedonian (En-Mk) and English-Albanian (En-Sq), yielding more than +8.3 BLEU points for all four translation directions.

arXiv:2009.06891공식 코드 · Papers with Code

Global-aware Beam Search for Neural Abstractive Summarization

This study develops a calibrated beam-based algorithm with awareness of the global attention distribution for neural abstractive summarization, aiming to improve the local optimality problem of the original beam search in a rigorous way. Specifically, a novel global protocol is proposed based on the attention distribution to stipulate how a global optimal hypothesis should attend to the source. A global scoring mechanism is then developed to regulate beam search to generate summaries in a near-global optimal fashion. This novel design enjoys a distinctive property, i.e., the global attention distribution could be predicted before inference, enabling step-wise improvements on the beam search through the global scoring mechanism. Extensive experiments on nine datasets show that the global (attention)-aware inference significantly improves state-of-the-art summarization models even using empirical hyper-parameters. The algorithm is also proven robust as it remains to generate meaningful texts with corrupted attention distributions. The codes and a comprehensive set of examples are available.

arXiv:2009.07253공식 코드 · Papers with Code

Autoregressive Knowledge Distillation through Imitation Learning

The performance of autoregressive models on natural language generation tasks has dramatically improved due to the adoption of deep, self-attentive architectures. However, these gains have come at the cost of hindering inference speed, making state-of-the-art models cumbersome to deploy in real-world, time-sensitive settings. We develop a compression technique for autoregressive models that is driven by an imitation learning perspective on knowledge distillation. The algorithm is designed to address the exposure bias problem. On prototypical language generation tasks such as translation and summarization, our method consistently outperforms other distillation algorithms, such as sequence-level knowledge distillation. Student models trained with our method attain 1.4 to 4.8 BLEU/ROUGE points higher than those trained from scratch, while increasing inference speed by up to 14 times in comparison to the teacher model.

arXiv:2009.07177공식 코드 · Papers with Code

Iterative Refinement in the Continuous Space for Non-Autoregressive Neural Machine Translation

We propose an efficient inference procedure for non-autoregressive machine translation that iteratively refines translation purely in the continuous space. Given a continuous latent variable model for machine translation (Shu et al., 2020), we train an inference network to approximate the gradient of the marginal log probability of the target sentence, using only the latent variable as input. This allows us to use gradient-based optimization to find the target sentence at inference time that approximately maximizes its marginal probability. As each refinement step only involves computation in the latent space of low dimensionality (we use 8 in our experiments), we avoid computational overhead incurred by existing non-autoregressive inference procedures that often refine in token space. We compare our approach to a recently proposed EM-like inference procedure (Shu et al., 2020) that optimizes in a hybrid space, consisting of both discrete and continuous variables. We evaluate our approach on WMT'14 En-De, WMT'16 Ro-En and IWSLT'16 De-En, and observe two advantages over the EM-like inference: (1) it is computationally efficient, i.e. each refinement step is twice as fast, and (2) it is more effective, resulting in higher marginal probabilities and BLEU scores with the same number of refinement steps. On WMT'14 En-De, for instance, our approach is able to decode 6.2 times faster than the autoregressive model with minimal degradation to translation quality (0.9 BLEU).

arXiv:2009.06451

Development of a Dataset and a Deep Learning Baseline Named Entity Recognizer for Three Low Resource Languages: Bhojpuri, Maithili and Magahi

In Natural Language Processing (NLP) pipelines, Named Entity Recognition (NER) is one of the preliminary problems, which marks proper nouns and other named entities such as Location, Person, Organization, Disease etc. Such entities, without a NER module, adversely affect the performance of a machine translation system. NER helps in overcoming this problem by recognising and handling such entities separately, although it can be useful in Information Extraction systems also. Bhojpuri, Maithili and Magahi are low resource languages, usually known as Purvanchal languages. This paper focuses on the development of a NER benchmark dataset for the Machine Translation systems developed to translate from these languages to Hindi by annotating parts of their available corpora. Bhojpuri, Maithili and Magahi corpora of sizes 228373, 157468 and 56190 tokens, respectively, were annotated using 22 entity labels. The annotation considers coarse-grained annotation labels followed by the tagset used in one of the Hindi NER datasets. We also report a Deep Learning based baseline that uses an LSTM-CNNs-CRF model. The lower baseline F1-scores from the NER tool obtained by using Conditional Random Fields models are 96.73 for Bhojpuri, 93.33 for Maithili and 95.04 for Magahi. The Deep Learning-based technique (LSTM-CNNs-CRF) achieved 96.25 for Bhojpuri, 93.33 for Maithili and 95.44 for Magahi.

arXiv:2009.05935

Combining Word and Character Vector Representation on Neural Machine Translation

This paper describes combinations of word vector representation and character vector representation in English-Indonesian neural machine translation (NMT). Six configurations of NMT models were built with different input vector representations: word-based, combination of word and character representation using bidirectional LSTM(bi-LSTM), combination of word and character representation using CNN, combination of word and character representation by combining bi-LSTM and CNN by three different vector operations: addition, pointwise multiplication, and averaging. The experiment results showed that NMT models with concatenation of word and character representation obtained BLEU score higher than baseline model, ranging from 9.14 points to 11.65 points, for all models that combining both word and character representation, except the model that combining word and character representation using both bi-LSTM and CNN by addition operation. The highest BLEU score achieved was 42.48 compared to the 30.83 of the baseline model.

arXiv:2009.05460

Robust Neural Machine Translation: Modeling Orthographic and Interpunctual Variation

Neural machine translation systems typically are trained on curated corpora and break when faced with non-standard orthography or punctuation. Resilience to spelling mistakes and typos, however, is crucial as machine translation systems are used to translate texts of informal origins, such as chat conversations, social media posts and web pages. We propose a simple generative noise model to generate adversarial examples of ten different types. We use these to augment machine translation systems' training data and show that, when tested on noisy data, systems trained using adversarial examples perform almost as well as when translating clean data, while baseline systems' performance drops by 2-3 BLEU points. To measure the robustness and noise invariance of machine translation systems' outputs, we use the average translation edit rate between the translation of the original sentence and its noised variants. Using this measure, we show that systems trained on adversarial examples on average yield 50% consistency improvements when compared to baselines trained on clean data.

arXiv:2009.04707

On Target Segmentation for Direct Speech Translation

Recent studies on direct speech translation show continuous improvements by means of data augmentation techniques and bigger deep learning models. While these methods are helping to close the gap between this new approach and the more traditional cascaded one, there are many incongruities among different studies that make it difficult to assess the state of the art. Surprisingly, one point of discussion is the segmentation of the target text. Character-level segmentation has been initially proposed to obtain an open vocabulary, but it results on long sequences and long training time. Then, subword-level segmentation became the state of the art in neural machine translation as it produces shorter sequences that reduce the training time, while being superior to word-level models. As such, recent works on speech translation started using target subwords despite the initial use of characters and some recent claims of better results at the character level. In this work, we perform an extensive comparison of the two methods on three benchmarks covering 8 language directions and multilingual training. Subword-level segmentation compares favorably in all settings, outperforming its character-level counterpart in a range of 1 to 3 BLEU points.

arXiv:2009.04087

Central Yup'ik and Machine Translation of Low-Resource Polysynthetic Languages

Machine translation tools do not yet exist for the Yup'ik language, a polysynthetic language spoken by around 8,000 people who live primarily in Southwest Alaska. We compiled a parallel text corpus for Yup'ik and English and developed a morphological parser for Yup'ik based on grammar rules. We trained a seq2seq neural machine translation model with attention to translate Yup'ik input into English. We then compared the influence of different tokenization methods, namely rule-based, unsupervised (byte pair encoding), and unsupervised morphological (Morfessor) parsing, on BLEU score accuracy for Yup'ik to English translation. We find that using tokenized input increases the translation accuracy compared to that of unparsed input. Although overall Morfessor did best with a vocabulary size of 30k, our first experiments show that BPE performed best with a reduced vocabulary size.

arXiv:2009.02358

Recent Trends in the Use of Deep Learning Models for Grammar Error Handling

Grammar error handling (GEH) is an important topic in natural language processing (NLP). GEH includes both grammar error detection and grammar error correction. Recent advances in computation systems have promoted the use of deep learning (DL) models for NLP problems such as GEH. In this survey we focus on two main DL approaches for GEH: neural machine translation models and editor models. We describe the three main stages of the pipeline for these models: data preparation, training, and inference. Additionally, we discuss different techniques to improve the performance of these models at each stage of the pipeline. We compare the performance of different models and conclude with proposed future directions.

arXiv:2009.02902

TransModality: An End2End Fusion Method with Transformer for Multimodal Sentiment Analysis

Multimodal sentiment analysis is an important research area that predicts speaker's sentiment tendency through features extracted from textual, visual and acoustic modalities. The central challenge is the fusion method of the multimodal information. A variety of fusion methods have been proposed, but few of them adopt end-to-end translation models to mine the subtle correlation between modalities. Enlightened by recent success of Transformer in the area of machine translation, we propose a new fusion method, TransModality, to address the task of multimodal sentiment analysis. We assume that translation between modalities contributes to a better joint representation of speaker's utterance. With Transformer, the learned features embody the information both from the source modality and the target modality. We validate our model on multiple multimodal datasets: CMU-MOSI, MELD, IEMOCAP. The experiments show that our proposed method achieves the state-of-the-art performance.

arXiv:2009.03116공식 코드 · Papers with Code

Why Not Simply Translate? A First Swedish Evaluation Benchmark for Semantic Similarity

This paper presents the first Swedish evaluation benchmark for textual semantic similarity. The benchmark is compiled by simply running the English STS-B dataset through the Google machine translation API. This paper discusses potential problems with using such a simple approach to compile a Swedish evaluation benchmark, including translation errors, vocabulary variation, and productive compounding. Despite some obvious problems with the resulting dataset, we use the benchmark to compare the majority of the currently existing Swedish text representations, demonstrating that native models outperform multilingual ones, and that simple bag of words performs remarkably well.

arXiv:2009.02016공식 코드 · Papers with Code

Dynamic Context-guided Capsule Network for Multimodal Machine Translation

Multimodal machine translation (MMT), which mainly focuses on enhancing text-only translation with visual features, has attracted considerable attention from both computer vision and natural language processing communities. Most current MMT models resort to attention mechanism, global context modeling or multimodal joint representation learning to utilize visual features. However, the attention mechanism lacks sufficient semantic interactions between modalities while the other two provide fixed visual context, which is unsuitable for modeling the observed variability when generating translation. To address the above issues, in this paper, we propose a novel Dynamic Context-guided Capsule Network (DCCN) for MMT. Specifically, at each timestep of decoding, we first employ the conventional source-target attention to produce a timestep-specific source-side context vector. Next, DCCN takes this vector as input and uses it to guide the iterative extraction of related visual features via a context-guided dynamic routing mechanism. Particularly, we represent the input image with global and regional visual features, we introduce two parallel DCCNs to model multimodal context vectors with visual features at different granularities. Finally, we obtain two multimodal context vectors, which are fused and incorporated into the decoder for the prediction of the target word. Experimental results on the Multi30K dataset of English-to-German and English-to-French translation demonstrate the superiority of DCCN. Our code is available on https://github.com/DeepLearnXMU/MM-DCCN.

arXiv:2009.01468

Modeling Global Body Configurations in American Sign Language

American Sign Language (ASL) is the fourth most commonly used language in the United States and is the language most commonly used by Deaf people in the United States and the English-speaking regions of Canada. Unfortunately, until recently, ASL received little research. This is due, in part, to its delayed recognition as a language until William C. Stokoe's publication in 1960. Limited data has been a long-standing obstacle to ASL research and computational modeling. The lack of large-scale datasets has prohibited many modern machine-learning techniques, such as Neural Machine Translation, from being applied to ASL. In addition, the modality required to capture sign language (i.e. video) is complex in natural settings (as one must deal with background noise, motion blur, and the curse of dimensionality). Finally, when compared with spoken languages, such as English, there has been limited research conducted into the linguistics of ASL. We realize a simplified version of Liddell and Johnson's Movement-Hold (MH) Model using a Probabilistic Graphical Model (PGM). We trained our model on ASLing, a dataset collected from three fluent ASL signers. We evaluate our PGM against other models to determine its ability to model ASL. Finally, we interpret various aspects of the PGM and draw conclusions about ASL phonetics. The main contributions of this paper are

arXiv:2009.00743

Bidirectional Attention Network for Monocular Depth Estimation

In this paper, we propose a Bidirectional Attention Network (BANet), an end-to-end framework for monocular depth estimation (MDE) that addresses the limitation of effectively integrating local and global information in convolutional neural networks. The structure of this mechanism derives from a strong conceptual foundation of neural machine translation, and presents a light-weight mechanism for adaptive control of computation similar to the dynamic nature of recurrent neural networks. We introduce bidirectional attention modules that utilize the feed-forward feature maps and incorporate the global context to filter out ambiguity. Extensive experiments reveal the high degree of capability of this bidirectional attention model over feed-forward baselines and other state-of-the-art methods for monocular depth estimation on two challenging datasets -- KITTI and DIODE. We show that our proposed approach either outperforms or performs at least on a par with the state-of-the-art monocular depth estimation methods with less memory and computational complexity.