ST-BERT: Cross-modal Language Model Pre-training For End-to-end Spoken Language Understanding

Language model pre-training has shown promising results in various downstream tasks. In this context, we introduce a cross-modal pre-trained language model, called Speech-Text BERT (ST-BERT), to tackle end-to-end spoken language understanding (E2E SLU) tasks. Taking phoneme posterior and subword-level text as an input, ST-BERT learns a contextualized cross-modal alignment via our two proposed pre-training tasks: Cross-modal Masked Language Modeling (CM-MLM) and Cross-modal Conditioned Language Modeling (CM-CLM). Experimental results on three benchmarks present that our approach is effective for various SLU datasets and shows a surprisingly marginal performance degradation even when 1% of the training data are available. Also, our method shows further SLU performance gain via domain-adaptive pre-training with domain-specific speech-text pair data.

Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs

Graph neural networks have shown superior performance in a wide range of applications providing a powerful representation of graph-structured data. Recent works show that the representation can be further improved by auxiliary tasks. However, the auxiliary tasks for heterogeneous graphs, which contain rich semantic information with various types of nodes and edges, have less explored in the literature. In this paper, to learn graph neural networks on heterogeneous graphs we propose a novel self-supervised auxiliary learning method using meta-paths, which are composite relations of multiple edge types. Our proposed method is learning to learn a primary task by predicting meta-paths as auxiliary tasks. This can be viewed as a type of meta-learning. The proposed method can identify an effective combination of auxiliary tasks and automatically balance them to improve the primary task. Our methods can be applied to any graph neural networks in a plug-in manner without manual labeling or additional data. The experiments demonstrate that the proposed method consistently improves the performance of link prediction and node classification on heterogeneous graphs.

Which Strategies Matter for Noisy Label Classification? Insight into Loss and Uncertainty

Label noise is a critical factor that degrades the generalization performance of deep neural networks, thus leading to severe issues in real-world problems. Existing studies have employed strategies based on either loss or uncertainty to address noisy labels, and ironically some strategies contradict each other: emphasizing or discarding uncertain samples or concentrating on high or low loss samples. To elucidate how opposing strategies can enhance model performance and offer insights into training with noisy labels, we present analytical results on how loss and uncertainty values of samples change throughout the training process. From the in-depth analysis, we design a new robust training method that emphasizes clean and informative samples, while minimizing the influence of noise using both loss and uncertainty. We demonstrate the effectiveness of our method with extensive experiments on synthetic and real-world datasets for various deep learning models. The results show that our method significantly outperforms other state-of-the-art methods and can be used generally regardless of neural network architectures.

CareCall: a Call-Based Active Monitoring Dialog Agent for Managing COVID-19 Pandemic

Tracking suspected cases of COVID-19 is crucial to suppressing the spread of COVID-19 pandemic. Active monitoring and proactive inspection are indispensable to mitigate COVID-19 spread, though these require considerable social and economic expense. To address this issue, we introduce CareCall, a call-based dialog agent which is deployed for active monitoring in Korea and Japan. We describe our system with a case study with statistics to show how the system works. Finally, we discuss a simple idea which uses CareCall to support proactive inspection.

Efficient Active Learning for Automatic Speech Recognition via Augmented Consistency Regularization

The cost of labeling transcriptions for large speech corpora becomes a bottleneck to maximally enjoy the potential capacity of deep neural network-based automatic speech recognition (ASR) models. Therefore, in this paper, we present a new training scheme that minimizes the labeling cost by adopting the concepts of semi-supervised learning (SSL) and active learning (AL) approaches and making a synergy from them. While AL studies only focus on selecting minimized the number of samples to be labeled with a criterion and taking advantage of such samples, we show that the training efficiency can be further improved by utilizing the unlabeled samples by sophisticatedly designing unsupervised loss that complements the unwanted behavior of supervised loss effectively. Our unsupervised loss is built on Consistency-Regularization (CR) approach, and we propose appropriate augmentation techniques to adopt CR in ASR field successfully. From the qualitative and quantitative experiments on the real-world dataset from deployed end-user voice assistant services, we show that the proposed methods can handle a large number of unlabeled speech data to achieve competitive model performance, with a sustainable amount of human labeling cost.

Slowing Down the Weight Norm Increase in Momentum-based Optimizers

Normalization techniques, such as batch normalization (BN), have led to significant improvements in deep neural network performances. Prior studies have analyzed the benefits of the resulting scale invariance of the weights for the gradient descent (GD) optimizers: it leads to a stabilized training due to the auto-tuning of step sizes. However, we show that, combined with the momentum-based algorithms, the scale invariance tends to induce an excessive growth of the weight norms. This in turn overly suppresses the effective step sizes during training, potentially leading to sub-optimal performances in deep neural networks. We analyze this phenomenon both theoretically and empirically. We propose a simple and effective solution: at each iteration of momentum-based GD optimizers (e.g. SGD or Adam) applied on scale-invariant weights (e.g. Conv weights preceding a BN layer), we remove the radial component (i.e. parallel to the weight vector) from the update vector. Intuitively, this operation prevents the unnecessary update along the radial direction that only increases the weight norm without contributing to the loss minimization. We verify that the modified optimizers SGDP and AdamP successfully regularize the norm growth and improve the performance of a broad set of models. Our experiments cover tasks including image classification and retrieval, object detection, robustness benchmarks, and audio classification. Source code is available at https://github.com/clovaai/AdamP.

Graphs, Entities, and Step Mixture

Existing approaches for graph neural networks commonly suffer from the oversmoothing issue, regardless of how neighborhoods are aggregated. Most methods also focus on transductive scenarios for fixed graphs, leading to poor generalization for unseen graphs. To address these issues, we propose a new graph neural network that considers both edge-based neighborhood relationships and node-based entity features, i.e. Graph Entities with Step Mixture via random walk (GESM). GESM employs a mixture of various steps through random walk to alleviate the oversmoothing problem, attention to dynamically reflect interrelations depending on node information, and structure-based regularization to enhance embedding representation. With intensive experiments, we show that the proposed GESM achieves state-of-the-art or comparable performances on eight benchmark graph datasets comprising transductive and inductive learning tasks. Furthermore, we empirically demonstrate the significance of considering global information.

ClovaCall: Korean Goal-Oriented Dialog Speech Corpus for Automatic Speech Recognition of Contact Centers

Automatic speech recognition (ASR) via call is essential for various applications, including AI for contact center (AICC) services. Despite the advancement of ASR, however, most publicly available call-based speech corpora such as Switchboard are old-fashioned. Also, most existing call corpora are in English and mainly focus on open domain dialog or general scenarios such as audiobooks. Here we introduce a new large-scale Korean call-based speech corpus under a goal-oriented dialog scenario from more than 11,000 people, i.e., ClovaCall corpus. ClovaCall includes approximately 60,000 pairs of a short sentence and its corresponding spoken utterance in a restaurant reservation domain. We validate the effectiveness of our dataset with intensive experiments using two standard ASR models. Furthermore, we release our ClovaCall dataset and baseline source codes to be available via https://github.com/ClovaAI/ClovaCall.