Recent work in multilingual translation advances translation quality surpassing bilingual baselines using deep transformer models with increased capacity.
During inference, BLT first generates a draft layout from the input and then iteratively refines it into a high-quality layout by masking out low-confident attributes.
Specifically, with the first attention function, Luna packs the input sequence into a sequence of fixed length.
Fully non-autoregressive neural machine translation (NAT) is proposed to simultaneously predict tokens with single forward of neural networks, which significantly reduces the inference latency at the expense of quality drop compared to the Transformer baseline.
In this work, we introduce a novel local autoregressive translation (LAT) mechanism into non-autoregressive translation (NAT) models so as to capture local dependencies among tar-get outputs.
Ranked #2 on Machine Translation on WMT2016 English-Romanian
Recently, pre-training contextualized encoders with language model (LM) objectives has been shown an effective semi-supervised method for structured prediction.
As an extension of this framework, we propose a novel method to train one shared Transformer network for multilingual machine translation with different layer selection posteriors for each language pair.
We introduce SCDE, a dataset to evaluate the performance of computational models through sentence prediction.
Ranked #3 on Question Answering on SCDE
In this work, we propose a new generative model that is capable of automatically decoupling global and local representations of images in an entirely unsupervised setting, by embedding a generative flow in the VAE framework to model the decoder.
Specifically, embeddings of entities and relationships are first decompressed to a more expressive and robust space by decompressing functions, then knowledge graph embedding models are trained in this new feature space.
Ranked #26 on Link Prediction on FB15k-237
Translation to or from low-resource languages LRLs poses challenges for machine translation in terms of both adequacy and fluency.
Flow-based generative models, conceptually attractive due to tractability of both the exact log-likelihood computation and latent-variable inference, and efficiency of both training and sampling, has led to a number of impressive empirical successes and spawned many advanced variants and theoretical investigations.
Ranked #3 on Image Generation on CelebA 256x256
In this work, we propose a method for neural dialogue response generation that allows not only generating semantically reasonable responses according to the dialogue history, but also explicitly controlling the sentiment of the response via sentiment labels.
Although Neural Machine Translation (NMT) models have advanced state-of-the-art performance in machine translation, they face problems like the inadequate translation.
Ranked #31 on Machine Translation on WMT2014 English-German
Mixture of Softmaxes (MoS) has been shown to be effective at addressing the expressiveness limitation of Softmax-based models.
Ranked #17 on Machine Translation on WMT2014 English-French
This paper describes methods for evaluating automatic speech recognition (ASR) systems in comparison with human perception results, using measures derived from linguistic distinctive features.
Mismatched transcriptions have been proposed as a mean to acquire probabilistic transcriptions from non-native speakers of a language. Prior work has demonstrated the value of these transcriptions by successfully adapting cross-lingual ASR systems for different tar-get languages.
Three consonant voicing classifiers were developed: (1) manually selected acoustic features anchored at a phonetic landmark, (2) MFCCs (either averaged across the segment or anchored at the landmark), and(3) acoustic features computed using a convolutional neural network (CNN).