A good neural sequence-to-sequence summarization model should have a strong encoder that can distill and memorize the important information from long input texts so that the decoder can generate salient summaries based on the encoder's memory. In this paper, we aim to improve the memorization capabilities of the encoder of a pointer-generator model by adding an additional 'closed-book' decoder without attention and pointer mechanisms. Such a decoder forces the encoder to be more selective in the information encoded in its memory state because the decoder can't rely on the extra information provided by the attention and possibly copy modules, and hence improves the entire model. On the CNN/Daily Mail dataset, our 2-decoder model outperforms the baseline significantly in terms of ROUGE and METEOR metrics, for both cross-entropy and reinforced setups (and on human evaluation). Moreover, our model also achieves higher scores in a test-only DUC-2002 generalizability setup. We further present a memory ability test, two saliency metrics, as well as several sanity-check ablations (based on fixed-encoder, gradient-flow cut, and model capacity) to prove that the encoder of our 2-decoder model does in fact learn stronger memory representations than the baseline encoder.
To improve the training efficiency of hierarchical recurrent models without compromising their performance, we propose a strategy named as `the lower the simpler', which is to simplify the baseline models by making the lower layers simpler than the upper layers. We carry out this strategy to simplify two typical hierarchical recurrent models, namely Hierarchical Recurrent Encoder-Decoder (HRED) and R-NET, whose basic building block is GRU. Specifically, we propose Scalar Gated Unit (SGU), which is a simplified variant of GRU, and use it to replace the GRUs at the middle layers of HRED and R-NET. Besides, we also use Fixed-size Ordinally-Forgetting Encoding (FOFE), which is an efficient encoding method without any trainable parameter, to replace the GRUs at the bottom layers of HRED and R-NET. The experimental results show that the simplified HRED and the simplified R-NET contain significantly less trainable parameters, consume significantly less training time, and achieve slightly better performance than their baseline models.
We investigate the task of distractor generation for multiple choice reading comprehension questions from examinations. In contrast to all previous works, we do not aim at preparing words or short phrases distractors, instead, we endeavor to generate longer and semantic-rich distractors which are closer to distractors in real reading comprehension from examinations. Taking a reading comprehension article, a pair of question and its correct option as input, our goal is to generate several distractors which are somehow related to the answer, consistent with the semantic context of the question and have some trace in the article. We propose a hierarchical encoder-decoder framework with static and dynamic attention mechanisms to tackle this task. Specifically, the dynamic attention can combine sentence-level and word-level attention varying at each recurrent time step to generate a more readable sequence. The static attention is to modulate the dynamic attention not to focus on question irrelevant sentences or sentences which contribute to the correct option. Our proposed framework outperforms several strong baselines on the first prepared distractor generation dataset of real reading comprehension questions. For human evaluation, compared with those distractors generated by baselines, our generated distractors are more functional to confuse the annotators.
Recent neural models for data-to-text generation are mostly based on data-driven end-to-end training over encoder-decoder networks. Even though the generated texts are mostly fluent and informative, they often generate descriptions that are not consistent with the input structured data. This is a critical issue especially in domains that require inference or calculations over raw data. In this paper, we attempt to improve the fidelity of neural data-to-text generation by utilizing pre-executed symbolic operations. We propose a framework called Operation-guided Attention-based sequence-to-sequence network (OpAtt), with a specifically designed gating mechanism as well as a quantization module for operation results to utilize information from pre-executed operations. Experiments on two sports datasets show our proposed method clearly improves the fidelity of the generated texts to the input structured data.
The encoder-decoder models for unsupervised sentence representation learning tend to discard the decoder after being trained on a large unlabelled corpus, since only the encoder is needed to map the input sentence into a vector representation. However, parameters learnt in the decoder also contain useful information about language. In order to utilise the decoder after learning, we present two types of decoding functions whose inverse can be easily derived without expensive inverse calculation. Therefore, the inverse of the decoding function serves as another encoder that produces sentence representations. We show that, with careful design of the decoding functions, the model learns good sentence representations, and the ensemble of the representations produced from the encoder and the inverse of the decoder demonstrate even better generalisation ability and solid transferability.
Recent neural machine translation (NMT) systems have been greatly improved by encoder-decoder models with attention mechanisms and sub-word units. However, important differences between languages with logographic and alphabetic writing systems have long been overlooked. This study focuses on these differences and uses a simple approach to improve the performance of NMT systems utilizing decomposed sub-character level information for logographic languages. Our results indicate that our approach not only improves the translation capabilities of NMT systems between Chinese and English, but also further improves NMT systems between Chinese and Japanese, because it utilizes the shared information brought by similar sub-character units.
Question Generation is the task of automatically creating questions from textual input. In this work we present a new Attentional Encoder--Decoder Recurrent Neural Network model for automatic question generation. Our model incorporates linguistic features and an additional sentence embedding to capture meaning at both sentence and word levels. The linguistic features are designed to capture information related to named entity recognition, word case, and entity coreference resolution. In addition our model uses a copying mechanism and a special answer signal that enables generation of numerous diverse questions on a given sentence. Our model achieves state of the art results of 19.98 Bleu_4 on a benchmark Question Generation dataset, outperforming all previously published results by a significant margin. A human evaluation also shows that these added features improve the quality of the generated questions.
Deep generative models have achieved remarkable success in various data domains, including images, time series, and natural languages. There remain, however, substantial challenges for combinatorial structures, including graphs. One of the key challenges lies in the difficulty of ensuring semantic validity in context. For examples, in molecular graphs, the number of bonding-electron pairs must not exceed the valence of an atom; whereas in protein interaction networks, two proteins may be connected only when they belong to the same or correlated gene ontology terms. These constraints are not easy to be incorporated into a generative model. In this work, we propose a regularization framework for variational autoencoders as a step toward semantic validity. We focus on the matrix representation of graphs and formulate penalty terms that regularize the output distribution of the decoder to encourage the satisfaction of validity constraints. Experimental results confirm a much higher likelihood of sampling valid graphs in our approach, compared with others reported in the literature.
As Deep Neural Networks (DNNs) are considered the state-of-the-art in many classification tasks, the question of their semantic generalizations has been raised. To address semantic interpretability of learned features, we introduce a novel idea of classification by re-generation based on variational autoencoder (VAE) in which a separate encoder-decoder pair of VAE is trained for each class. Moreover, the proposed architecture overcomes the scalability issue in current DNN networks as there is no need to re-train the whole network with the addition of new classes and it can be done for each class separately. We also introduce a criterion based on Kullback-Leibler divergence to reject doubtful examples. This rejection criterion should improve the trust in the obtained results and can be further exploited to reject adversarial examples.
Face completion is a challenging generation task because it requires generating visually pleasing new pixels that are semantically consistent with the unmasked face region. This paper proposes a geometry-aware Face Completion and Editing NETwork (FCENet) by systematically studying facial geometry from the unmasked region. Firstly, a facial geometry estimator is learned to estimate facial landmark heatmaps and parsing maps from the unmasked face image. Then, an encoder-decoder structure generator serves to complete a face image and disentangle its mask areas conditioned on both the masked face image and the estimated facial geometry images. Besides, since low-rank property exists in manually labeled masks, a low-rank regularization term is imposed on the disentangled masks, enforcing our completion network to manage occlusion area with various shape and size. Furthermore, our network can generate diverse results from the same masked input by modifying estimated facial geometry, which provides a flexible mean to edit the completed face appearance. Extensive experimental results qualitatively and quantitatively demonstrate that our network is able to generate visually pleasing face completion results and edit face attributes as well.
Neural machine translation (NMT) systems operate primarily on words (or sub-words), ignoring lower-level patterns of morphology. We present a character-aware decoder designed to capture such patterns when translating into morphologically rich languages. We achieve character-awareness by augmenting both the softmax and embedding layers of an attention-based encoder-decoder model with convolutional neural networks that operate on the spelling of a word. To investigate performance on a wide variety of morphological phenomena, we translate English into 14 typologically diverse target languages using the TED multi-target dataset. In this low-resource setting, the character-aware decoder provides consistent improvements with BLEU score gains of up to $+3.05$. In addition, we analyze the relationship between the gains obtained and properties of the target language and find evidence that our model does indeed exploit morphological patterns.
The addition of syntax-aware decoding in Neural Machine Translation (NMT) systems requires an effective tree-structured neural network, a syntax-aware attention model and a language generation model that is sensitive to sentence structure. We exploit a top-down tree-structured model called DRNN (Doubly-Recurrent Neural Networks) first proposed by Alvarez-Melis and Jaakola (2017) to create an NMT model called Seq2DRNN that combines a sequential encoder with tree-structured decoding augmented with a syntax-aware attention model. Unlike previous approaches to syntax-based NMT which use dependency parsing models our method uses constituency parsing which we argue provides useful information for translation. In addition, we use the syntactic structure of the sentence to add new connections to the tree-structured decoder neural network (Seq2DRNN+SynC). We compare our NMT model with sequential and state of the art syntax-based NMT models and show that our model produces more fluent translations with better reordering. Since our model is capable of doing translation and constituency parsing at the same time we also compare our parsing accuracy against other neural parsing models.
We propose a joint source and channel coding (JSCC) technique for wireless image transmission that does not rely on explicit codes for either compression or error correction; instead, it directly maps the image pixel values to the complex-valued channel input symbols. We parameterize the encoder and decoder functions by two convolutional neural networks (CNNs), which are trained jointly, and can be considered as an autoencoder with a non-trainable layer in the middle that represents the noisy communication channel. Our results show that the proposed deep JSCC scheme outperforms digital transmission concatenating JPEG or JPEG2000 compression with a capacity achieving channel code at low signal-to-noise ratio (SNR) and channel bandwidth values in the presence of additive white Gaussian noise (AWGN). More strikingly, deep JSCC does not suffer from the ``cliff effect'', and it provides a graceful performance degradation as the channel SNR varies with respect to the SNR value assumed during training. In the case of a slow Rayleigh fading channel, deep JSCC learns noise resilient coded representations and significantly outperforms separation-based digital communication at all SNR and channel bandwidth values.
In this paper, we present Adaptive Computation Steps (ACS) algo-rithm, which enables end-to-end speech recognition models to dy-namically decide how many frames should be processed to predict a linguistic output. The model that applies ACS algorithm follows the encoder-decoder framework, while unlike the attention-based mod-els, it produces alignments independently at the encoder side using the correlation between adjacent frames. Thus, predictions can be made as soon as sufficient acoustic information is received, which makes the model applicable in online cases. Besides, a small change is made to the decoding stage of the encoder-decoder framework, which allows the prediction to exploit bidirectional contexts. We verify the ACS algorithm on a Mandarin speech corpus AIShell-1, and it achieves a 31.2% CER in the online occasion, compared to the 32.4% CER of the attention-based model. To fully demonstrate the advantage of ACS algorithm, offline experiments are conducted, in which our ACS model achieves an 18.7% CER, outperforming the attention-based counterpart with the CER of 22.0%.
Inspired by the recent success of fully convolutional networks (FCN) in semantic segmentation, we propose a deep smoke segmentation network to infer high quality segmentation masks from blurry smoke images. To overcome large variations in texture, color and shape of smoke appearance, we divide the proposed network into a coarse path and a fine path. The first path is an encoder-decoder FCN with skip structures, which extracts global context information of smoke and accordingly generates a coarse segmentation mask. To retain fine spatial details of smoke, the second path is also designed as an encoder-decoder FCN with skip structures, but it is shallower than the first path network. Finally, we propose a very small network containing only add, convolution and activation layers to fuse the results of the two paths. Thus, we can easily train the proposed network end to end for simultaneous optimization of network parameters. To avoid the difficulty in manually labelling fuzzy smoke objects, we propose a method to generate synthetic smoke images. According to results of our deep segmentation method, we can easily and accurately perform smoke detection from videos. Experiments on three synthetic smoke datasets and a realistic smoke dataset show that our method achieves much better performance than state-of-the-art segmentation algorithms based on FCNs. Test results of our method on videos are also appealing.
Neural Machine Translation (NMT) can be improved by including document-level contextual information. For this purpose, we propose a hierarchical attention model to capture the context in a structured and dynamic manner. The model is integrated in the original NMT architecture as another level of abstraction, conditioning on the NMT model's own previous hidden states. Experiments show that hierarchical attention significantly improves the BLEU score over a strong NMT baseline with the state-of-the-art in context-aware methods, and that both the encoder and decoder benefit from context in complementary ways.
This paper documents the Team Copenhagen system which placed first in the CoNLL--SIGMORPHON 2018 shared task on universal morphological reinflection, Task 2 with an overall accuracy of 49.87. Task 2 focuses on morphological inflection in context: generating an inflected word form, given the lemma of the word and the context it occurs in. Previous SIGMORPHON shared tasks have focused on context-agnostic inflection---the "inflection in context" task was introduced this year. We approach this with an encoder-decoder architecture over character sequences with three core innovations, all contributing to an improvement in performance: (1) a wide context window; (2) a multi-task learning approach with the auxiliary task of MSD prediction; (3) training models in a multilingual fashion.
Contemporary deep learning based medical image segmentation algorithms require hours of annotation labor by domain experts. These data hungry deep models perform sub-optimally in the presence of limited amount of labeled data. In this paper, we present a data efficient learning framework using the recent concept of Generative Adversarial Networks; this allows a deep neural network to perform significantly better than its fully supervised counterpart in low annotation regime. The proposed method is an extension of our previous work with the addition of a new unsupervised adversarial loss and a structured prediction based architecture. To the best of our knowledge, this work is the first demonstration of an adversarial framework based structured prediction model for medical image segmentation. Though generic, we apply our method for segmentation of blood vessels in retinal fundus images. We experiment with extreme low annotation budget (0.8 - 1.6% of contemporary annotation size). On DRIVE and STARE datasets, the proposed method outperforms our previous method and other fully supervised benchmark models by significant margins especially with very low number of annotated examples. In addition, our systematic ablation studies suggest some key recipes for successfully training GAN based semi-supervised algorithms with an encoder-decoder style network architecture.
Word sense disambiguation (WSD) is a well researched problem in computational linguistics. Different research works have approached this problem in different ways. Some state of the art results that have been achieved for this problem are by supervised models in terms of accuracy, but they often fall behind flexible knowledge-based solutions which use engineered features as well as human annotators to disambiguate every target word. This work focuses on bridging this gap using neural sequence models incorporating the well-known attention mechanism. The main gist of our work is to combine multiple attentions on different linguistic features through weights and to provide a unified framework for doing this. This weighted attention allows the model to easily disambiguate the sense of an ambiguous word by attending over a suitable portion of a sentence. Our extensive experiments show that multiple attention enables a more versatile encoder-decoder model leading to state of the art results.
In this paper, a neural network-aided bit-interleaved coded modulation (NN-BICM) receiver is designed to mitigate the nonlinear clipping distortion in the LDPC coded direct currentbiased optical orthogonal frequency division multiplexing (DCOOFDM) systems. Taking the cross-entropy as loss function, a feed forward network is trained by backpropagation algorithm to output the condition probability through the softmax activation function, thereby assisting in a modified log-likelihood ratio (LLR) improvement. To reduce the complexity, this feed-forward network simplifies the input layer with a single symbol and the corresponding Gaussian variance instead of focusing on the inter-carrier interference between multiple subcarriers. On the basis of the neural network-aided BICM with Gray labelling, we propose a novel stacked network architecture of the bitinterleaved coded modulation with iterative decoding (NN-BICMID). Its performance has been improved further by calculating the condition probability with the aid of a priori probability that derived from the extrinsic LLRs in the LDPC decoder at the last iteration, at the expense of customizing neural network detectors at each iteration time separately. Utilizing the optimal DC bias as the midpoint of the dynamic region, the simulation results demonstrate that both the NN-BICM and NN-BICM-ID schemes achieve noticeable performance gains than other counterparts, in which the NN-BICM-ID clearly outperforms NN-BICM with various modulation and coding schemes.