United States

Non-autoregressive neural machine translation (NAT) generates each target word in parallel and has achieved promising inference acceleration. However, existing NAT models still have a big gap in translation quality compared to autoregressive neural machine translation models due to the multimodality problem: the target words may come from multiple feasible translations. To address this problem, we propose a novel NAT framework ReorderNAT which explicitly models the reordering information to guide the decoding of NAT. Specially, ReorderNAT utilizes deterministic and non-deterministic decoding strategies that leverage reordering information as a proxy for the final translation to encourage the decoder to choose words belonging to the same translation. Experimental results on various widely-used datasets show that our proposed model achieves better performance compared to most existing NAT models, and even achieves comparable translation quality as autoregressive translation models with a significant speedup.

AAAI 2021

Guiding Non-Autoregressive Neural Machine Translation Decoding with Reordering Information

2021 Annual Conference of the North American Chapter of the Association for Computational Linguistics.

**Whova App** 
Stay in touch with your fellow conference attendees via the [Whova App](https://whova.com/portal/webapp/nacon_202106/)

**Conference Structure**
https://2021.naacl.org/blog/conference-structure/

**Walkthrough video of how to NAACL 2021** 

Please take a moment to view this video explaining how to navigate the platform, attend sessions network with other attendees. 


<figure class="video_container">
  <iframe src="https://screencast-o-matic.com/watch/crhwbGVh3vx?v=6&ff=1&title=0&controls=1" width=640  height=350  frameborder="0" allowfullscreen="true"> </iframe>
</figure>

<br>
<br>

NAACL 2021

technical paper

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

Despite having impressive vision-language (VL) pretraining with BERT-based encoder for VL understanding, the pretraining of a universal encoder-decoder for both VL understanding and generation remains challenging. The difficulty originates from the inherently different peculiarities of the two disciplines, e.g., VL understanding tasks capitalize on the unrestricted message passing across modalities, while generation tasks only employ visual-to-textual message passing. In this paper, we start with a two-stream decoupled design of encoder-decoder structure, in which two decoupled cross-modal encoder and decoder are involved to separately perform each type of proxy tasks, for simultaneous VL understanding and generation pretraining. Moreover, for VL pretraining, the dominant way is to replace some input visual/word tokens with mask tokens and enforce the multi-modal encoder/decoder to reconstruct the original tokens, but no mask token is involved when fine-tuning on downstream tasks. As an alternative, we propose a primary scheduled sampling strategy that elegantly mitigates such discrepancy via pretraining encoder-decoder in a two-pass manner. Extensive experiments demonstrate the compelling generalizability of our pretrained encoder-decoder by fine-tuning on four VL understanding and generation downstream tasks. Source code is available at https://github.com/YehLi/TDEN.

Scheduled Sampling in Vision-Language Pretraining with Decoupled Encoder-Decoder Network

Intelligent transportation management requires not only statistical information on users' mobility patterns, but also knowledge of their corresponding transportation modes. While GPS trajectories can be readily obtained from GPS sensors found in modern smartphones and vehicles, these massive geospatial data are neither automatically annotated nor segmented by transportation mode, subsequently complicating transportation mode identification. In addition, predictive uncertainty caused by the learned model parameters or variable noise in GPS sensor readings typically remains unaccounted for. To jointly address the above issues, we propose a Bayesian deep learning framework for unsupervised GPS trajectory segmentation. After unlabeled GPS trajectories are preprocessed into sequences of motion features, they are used in unsupervised training of a channel-calibrated temporal convolutional neural network for timestep-level transportation mode identification. At test time, we approximate variational inference via Monte Carlo dropout sampling, leveraging the mean and variance of the predicted distributions to classify each input timestep and estimate its predictive uncertainty, respectively. The proposed approach outperforms both its non-Bayesian variant and established GPS trajectory segmentation baselines on Microsoft's Geolife dataset without using any labels.

Capturing Uncertainty in Unsupervised GPS Trajectory Segmentation Using Bayesian Deep Learning

To simultaneously translate a source language into multiple different target languages is one of the most common scenarios of multilingual translation. However, existing methods cannot make full use of translation model information during decoding, such as intra-lingual and inter-lingual future information, and therefore may suffer from some issues like the unbalanced outputs. In this paper, we present a new approach for synchronous interactive multilingual neural machine translation (SimNMT), which predicts each target language output simultaneously and interactively using historical and future information of all target languages. Specifically, we first propose a synchronous cross-interactive decoder in which generation of each target output does not only depend on its generated sequences, but also relies on its future information, as well as history and future contexts of other target languages. Then, we present a new interactive multilingual beam search algorithm that enables synchronous interactive decoding of all target languages in a single model. We take two target languages as an example to illustrate and evaluate the proposed SimNMT model on IWSLT datasets. The experimental results demonstrate that our method achieves significant improvements over several advanced NMT and MNMT models.

Synchronous Interactive Decoding for Multilingual Neural Machine Translation

Differentiable architecture search (DAS) has made great progress in searching for  high-performance architectures with reduced computational cost. However, DAS-based methods mainly focus on searching for a repeatable cell structure, which is then stacked sequentially in multiple stages to form the networks. This configuration significantly reduces the search space, and ignores the importance of connections between the cells. To overcome this limitation, in this paper, we propose a Hierarchical Differentiable Architecture Search (H-DAS) that performs architecture search both at the cell level and at the stage level. Specifically, the cell-level search space is relaxed so that the networks can learn stage-specific cell structures. For the stage-level search, we systematically study the architectures of stages, including the number of cells in each stage and the connections between the cells. Based on insightful observations, we design several search rules and losses, and mange to search for  better stage-level architectures. Such hierarchical search space greatly improves the performance of the networks without introducing expensive search cost. Extensive experiments on CIFAR10 and ImageNet demonstrate the effectiveness of the proposed H-DAS. Moreover, the searched stage-level architectures can be combined with the cell structures searched by existing DAS methods to further boost the performance. Code is available at: https://github.com/msight-tech/research-HDAS

Unchain the Search Space with Hierarchical Differentiable Architecture Search

Previous work has shown the effectiveness of using event representations for tasks such as script event prediction and stock market prediction. It is however still challenging to learn the subtle semantic differences between events based solely on textual descriptions of events often represented as (subject, predicate, object) triples. As an alternative, images offer a more intuitive way of understanding event semantics. We observe that event described in text and in images show different abstraction levels and therefore should be projected onto heterogeneous embedding spaces, as opposed to what have been done in previous approaches which project signals from different modalities onto a homogeneous space. In this paper, we propose a Multimodal Event Representation Learning framework (MERL) to learn event representations based on both text and image modalities simultaneously. Event textual triples are projected as Gaussian density embeddings by a dual-path Gaussian triple encoder, while event images are projected as point embeddings by a visual event component-aware image encoder. Moreover, a novel score function motivated by statistical hypothesis testing is introduced to coordinate two embedding spaces. Experiments are conducted on various multimodal event-related tasks and results show that MERL outperforms a number of unimodal and multimodal baselines, demonstrating the effectiveness of the proposed framework.

MERL: Multimodal Event Representation Learning in Heterogeneous Embedding Spaces

Lexical taxonomies, a special kind of knowledge graph, are essential for natural language understanding. This paper studies the problem of lexical taxonomy embedding. Most existing graph embedding methods are difficult to apply to lexical taxonomies since 1) they ignore implicit but important information, namely, sibling relations, which are not explicitly mentioned in lexical taxonomies and 2) there are lots of polysemous terms in lexical taxonomies. In this paper, we propose a novel method for lexical taxonomy embedding. This method optimizes an objective function that models both hyponym-hypernym relations and sibling relations. A term-level attention mechanism and a random walk based metric are then proposed to assist the modeling of these two kinds of relations, respectively. Finally, a novel training method based on curriculum learning is proposed. We conduct extensive experiments on two tasks to show that our approach outperforms other embedding methods and we use the learned term embeddings to enhance the performance of the state-of-the-art models that are based on BERT and RoBERTa on text classification.

Learning Term Embeddings for Lexical Taxonomies

Collecting supporting evidence from large corpora of text (e.g., Wikipedia) is of great challenge for open-domain Question Answering (QA). Especially, for multi-hop open-domain QA, scattered evidence pieces are required to be gathered together to support the answer extraction. In this paper, we propose a new retrieval target, hop, to collect the hidden reasoning evidence from Wikipedia for complex question answering. Specifically, the hop in this paper is defined as the combination of a hyperlink and the corresponding outbound link document. The hyperlink is encoded as the mention embedding which models the structured knowledge of how the outbound link entity is mentioned in the textual context, and the corresponding outbound link document is encoded as the document embedding representing the unstructured knowledge within it. Accordingly, we build HopRetriever which retrieves hops over Wikipedia to answer complex questions. Experiments on the HotpotQA dataset demonstrate that HopRetriever outperforms previously published evidence retrieval methods by large margins. Moreover, our approach also yields quantifiable interpretations of the evidence collection process.

HopRetriever: Retrieve Hops over Wikipedia to Answer Complex Questions

For prediction of interacting agents' trajectories,  we propose an end-to-end trainable architecture that hybridizes neural nets with game-theoretic reasoning, has interpretable intermediate representations, and transfers to downstream decision making. It uses a net that reveals preferences from the agents' past joint trajectory, and a differentiable implicit layer that maps these preferences to local Nash equilibria, forming the modes of the predicted future trajectory. Additionally, it learns an equilibrium refinement concept. For tractability, we introduce a new class of continuous potential games and an equilibrium-separating partition of the action space. We provide theoretical results for explicit gradients and soundness. In experiments, we evaluate our approach on two real-world data sets, where we predict highway drivers' merging trajectories, and on a simple decision-making transfer task.

Learning Game-Theoretic Models of Multiagent Trajectories Using Implicit Layers

It is well-known that second-order optimizer can accelerate the training of deep neural networks, however, the huge computation cost of second-order optimization makes it impractical to apply in real practice. In order to reduce the cost, many methods have been proposed to approximate a second-order matrix. Inspired by KFAC, we propose a novel Trace-based Hardware-driven layer-ORiented Natural Gradient Descent Computation method, called THOR, to make the second-order optimization applicable in the real application models. Specifically, we gradually increase the update interval and use the matrix trace to determine which blocks of Fisher Information Matrix (FIM) need to be updated. Moreover, by resorting the power of hardware, we have designed a Hardware-driven approximation method for computing FIM to achieve better performance. To demonstrate the effectiveness of THOR, we have conducted extensive experiments. The results show that training ResNet-50 on ImageNet with THOR only takes 66.7 minutes to achieve a top-1 accuracy of 75.9 % under an 8 Ascend 910 environment with MindSpore, a new deep learning computing framework. Moreover, with more computational resources, THOR can only takes 2.7 minutes to 75.9 % with 256 Ascend 910.

THOR, Trace-Based Hardware-Driven Layer-Oriented Natural Gradient Descent Computation

A well known result is that, given a consistent heuristic and no other source of information, A* does expand a minimal number of nodes up to tie-breaking. We extend this analysis for A* with dominance pruning, which exploits a dominance relation to eliminate some nodes during the search. We show that the expansion order of A* is not necessarily optimally efficient when considering dominance pruning with arbitrary dominance relations, but it remains optimally efficient under certain restrictions for the heuristic and dominance relation.

Downloads

Next from AAAI 2021

Scheduled Sampling in Vision-Language Pretraining with Decoupled Encoder-Decoder Network

Similar lecture

Multi-Task Learning with Shared Encoder for Non-Autoregressive Machine Translation

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Scheduled Sampling in Vision-Language Pretraining with Decoupled Encoder-Decoder Network

Similar lecture

Multi-Task Learning with Shared Encoder for Non-Autoregressive Machine Translation

Downloads