EMNLP 2020 Papers

Disfluency detection is usually an intermediate step between an automatic speech recognition (ASR) system and a downstream task. By contrast, this paper aims to investigate the task of end-to-end speech recognition and disfluency removal. We specifically explore whether it is possible to train an ASR model to directly map disfluent speech into fluent transcripts, without relying on a separate disfluency detection model. We show that end-to-end models do learn to directly generate fluent transcripts; however, their performance is slightly worse than a baseline pipeline approach consisting of an ASR system and a specialized disfluency detection model. We also propose two new metrics for evaluating integrated ASR and disfluency removal models. The findings of this paper can serve as a benchmark for further research on the task of end-to-end speech recognition and disfluency removal in the future.

Deep neural network-based pretraining methods have achieved impressive results in many natural language processing tasks including text classification. However, their applicability to large-scale text classification with numerous categories (e.g., several thousands) is yet to be well-studied, where the training data is insufficient and skewed in terms of categories. In addition, existing pretraining methods usually involve excessive computation and memory overheads. In this paper, we develop a novel multi-pretraining framework for large-scale text classification. This multi-pretraining framework includes both a self-supervised pretraining and a weakly supervised pretraining. We newly introduce an out-of-context words detection task on the unlabeled data as the self-supervised pretraining. It captures the topic-consistency of words used in sentences, which is proven to be useful for text classification. In addition, we propose a weakly supervised pretraining, where labels for text classification are obtained automatically from an existing approach. Experimental results clearly show that both pretraining approaches are effective for large-scale text classification task. The proposed scheme exhibits significant improvements as much as 3.8% in terms of macro-averaging F1-score over strong pretraining methods, while being computationally efficient.

Chinese spelling check is a challenging task due to the characteristics of the Chinese language, such as the large character set, no word boundary, and short word length. On the one hand, most of the previous works only consider corrections with similar character pronunciation or shape, failing to correct visually and phonologically irrelevant typos. On the other hand, pipeline-style architectures are widely adopted to deal with different types of spelling errors in individual modules, which is difficult to optimize. In order to handle these issues, in this work, 1) we extend the traditional confusion sets with semantical candidates to cover different types of errors; 2) we propose a chunk-based framework to correct single-character and multi-character word errors uniformly; and 3) we adopt a global optimization strategy to enable a sentence-level correction selection. The experimental results show that the proposed approach achieves a new state-of-the-art performance on three benchmark datasets, as well as an optical character recognition dataset.

To model diverse responses for a given post, one promising way is to introduce a latent variable into Seq2Seq models. The latent variable is supposed to capture the discourse-level information and encourage the informativeness of target responses. However, such discourse-level information is often too coarse for the decoder to be utilized. To tackle it, our idea is to transform the coarse-grained discourse-level information into fine-grained word-level information. Specifically, we firstly measure the semantic concentration of corresponding target response on the post words by introducing a fine-grained focus signal. Then, we propose a focus-constrained attention mechanism to take full advantage of focus in well aligning the input to the target response. The experimental results demonstrate that by exploiting the fine-grained signal, our model can generate more diverse and informative responses compared with several state-of-the-art models.

While Iterative Back-Translation and Dual Learning effectively incorporate monolingual training data in neural machine translation, they use different objectives and heuristic gradient approximation strategies, and have not been extensively compared. We introduce a novel dual reconstruction objective that provides a unified view of Iterative Back-Translation and Dual Learning. It motivates a theoretical analysis and controlled empirical study on German-English and Turkish-English tasks, which both suggest that Iterative Back-Translation is more effective than Dual Learning despite its relative simplicity.

Data programming aims to reduce the cost of curating training data by encoding domain knowledge as labeling functions over source data. As such it not only requires domain expertise but also programming experience, a skill that many subject matter experts lack. Additionally, generating functions by enumerating rules is not only time consuming but also inherently difficult, even for people with programming experience. In this paper we introduce Ruler, an interactive system that synthesizes labeling rules using span-level interactive demonstrations over document examples. Ruler is a first-of-a-kind implementation of data programming by demonstration (DPBD). This new framework aims to relieve users from the burden of writing labeling functions, enabling them to focus on higher-level semantic analysis, such as identifying relevant signals for the labeling task. We compare Ruler with conventional data programming through a user study conducted with 10 data scientists who were asked to create labeling functions for sentiment and spam classification tasks. Results show Ruler is easier to learn and to use, and that it offers higher overall user-satisfaction while providing model performances comparable to those achieved by conventional data programming.

The successful application of neural methods to machine translation has realized huge quality advances for the community. With these improvements, many have noted outstanding challenges, including the modeling and treatment of gendered language. While previous studies have identified issues using synthetic examples, we develop a novel technique to mine examples from real world data to explore challenges for deployed systems. We use our method to compile an evaluation benchmark spanning examples for four languages from three language families, which we publicly release to facilitate research. The examples in our benchmark expose where model representations are gendered, and the unintended consequences these gendered representations can have in downstream application.

As an important research topic, customer service dialogue generation tends to generate generic seller responses by leveraging current dialogue information. In this study, we propose a novel and extensible dialogue generation method by leveraging sellers’ historical dialogue information, which can be both accessible and informative. By utilizing innovative historical dialogue representation learning and historical dialogue selection mechanism, the proposed model is capable of detecting most related responses from sellers’ historical dialogues, which can further enhance the current dialogue generation quality. Unlike prior dialogue generation efforts, we treat each seller’s historical dialogues as a list of Customer-Seller utterance pairs and allow the model to measure their different importance, and copy words directly from most relevant pairs. Extensive experimental results show that the proposed approach can generate high-quality responses that cater to specific sellers’ characteristics and exhibit consistent superiority over baselines on a real-world multi-turn customer service dialogue dataset.

We introduce DropHead, a structured dropout method specifically designed for regularizing the multi-head attention mechanism which is a key component of transformer. In contrast to the conventional dropout mechanism which randomly drops units or connections, DropHead drops entire attention heads during training to prevent the multi-head attention model from being dominated by a small portion of attention heads. It can help reduce the risk of overfitting and allow the models to better benefit from the multi-head attention. Given the interaction between multi-headedness and training dynamics, we further propose a novel dropout rate scheduler to adjust the dropout rate of DropHead throughout training, which results in a better regularization effect. Experimental results demonstrate that our proposed approach can improve transformer models by 0.9 BLEU score on WMT14 En-De translation task and around 1.0 accuracy for various text classification tasks.

In this paper, we study a new task of synonym expansion using transitivity, and propose a novel approach named SynET, which considers both the contexts of two given synonym pairs. It introduces an auxiliary task to reduce the impact of noisy sentences, and proposes a Multi-Perspective Entity Matching Network to match entities from multiple perspectives. Extensive experiments on a real-world dataset show the effectiveness of our approach.

Automatic medical image report generation has drawn growing attention due to its potential to alleviate radiologists’ workload. Existing work on report generation often trains encoder-decoder networks to generate complete reports. However, such models are affected by data bias (e.g. label imbalance) and face common issues inherent in text generation models (e.g. repetition). In this work, we focus on reporting abnormal findings on radiology images; instead of training on complete radiology reports, we propose a method to identify abnormal findings from the reports in addition to grouping them with unsupervised clustering and minimal rules. We formulate the task as cross-modal retrieval and propose Conditional Visual-Semantic Embeddings to align images and fine-grained abnormal findings in a joint embedding space. We demonstrate that our method is able to retrieve abnormal findings and outperforms existing generation models on both clinical correctness and text generation metrics.

Exposing diverse subword segmentations to neural machine translation (NMT) models often improves the robustness of machine translation as NMT models can experience various subword candidates. However, the diversification of subword segmentations mostly relies on the pre-trained subword language models from which erroneous segmentations of unseen words are less likely to be sampled. In this paper, we present adversarial subword regularization (ADVSR) to study whether gradient signals during training can be a substitute criterion for exposing diverse subword segmentations. We experimentally show that our model-based adversarial samples effectively encourage NMT models to be less sensitive to segmentation errors and improve the performance of NMT models in low-resource and out-domain datasets.

User modeling is critical for many personalized web services. Many existing methods model users based on their behaviors and the labeled data of target tasks. However, these methods cannot exploit useful information in unlabeled user behavior data, and their performance may be not optimal when labeled data is scarce. Motivated by pre-trained language models which are pre-trained on large-scale unlabeled corpus to empower many downstream tasks, in this paper we propose to pre-train user models from large-scale unlabeled user behaviors data. We propose two self-supervision tasks for user model pre-training. The first one is masked behavior prediction, which can model the relatedness between historical behaviors. The second one is next K behavior prediction, which can model the relatedness between past and future behaviors. The pre-trained user models are finetuned in downstream tasks to learn task-specific user representations. Experimental results on two real-world datasets validate the effectiveness of our proposed user model pre-training method.

Image captioning systems need to produce texts that are not only true but also relevant in that they are properly aligned with the current issues. For instance, in a newspaper article about a sports event, a caption that not only identifies the player in a picture but also comments on their ethnicity could create unwanted reader reactions. To address this, we propose Issue-Sensitive Image Captioning (ISIC). In ISIC, the captioner is given a target image and an issue, which is a set of images partitioned in a way that specifies what information is relevant. For the sports article, we could construct a partition that places images into equivalence classes based on player position. To model this task, we use an extension of the Rational Speech Acts model. Our extension is built on top of state-of-the-art pretrained neural image captioners and explicitly uses image partitions to control caption generation. In both automatic and human evaluations, we show that these models generate captions that are descriptive and issue-sensitive. Finally, we show how ISIC can complement and enrich the related task of Visual Question Answering.

Learning representations of spatial references in natural language is a key challenge in tasks like autonomous navigation and robotic manipulation. Recent work has investigated various neural architectures for learning multi-modal representations for spatial concepts. However, the lack of explicit reasoning over entities makes such approaches vulnerable to noise in input text or state observations. In this paper, we develop effective models for understanding spatial references in text that are robust and interpretable, without sacrificing performance. We design a text-conditioned relation network whose parameters are dynamically computed with a cross-modal attention module to capture fine-grained spatial relations between entities. This design choice provides interpretability of learned intermediate outputs. Experiments across three tasks demonstrate that our model achieves superior performance, with a 17% improvement in predicting goal locations and a 15% improvement in robustness compared to state-of-the-art systems.

Question answering (QA) tasks have been posed using a variety of formats, such as extractive span selection, multiple choice, etc. This has led to format-specialized models, and even to an implicit division in the QA community. We argue that such boundaries are artificial and perhaps unnecessary, given the reasoning abilities we seek to teach are not governed by the format. As evidence, we use the latest advances in language modeling to build a single pre-trained QA model, UNIFIEDQA, that performs well across 19 QA datasets spanning 4 diverse formats. UNIFIEDQA performs on par with 8 different models that were trained on individual datasets themselves. Even when faced with 12 unseen datasets of observed formats, UNIFIEDQA performs surprisingly well, showing strong generalization from its outof-format training data. Finally, simply finetuning this pre trained QA model into specialized models results in a new state of the art on 10 factoid and commonsense question answering datasets, establishing UNIFIEDQA as a strong starting point for building QA systems.

We study relationships between spoken language and co-speech gestures in context of two key challenges. First, distributions of text and gestures are inherently skewed making it important to model the long tail. Second, gesture predictions are made at a subword level, making it important to learn relationships between language and acoustic cues. We introduce AISLe, which combines adversarial learning with importance sampling to strike a balance between precision and coverage. We propose the use of a multimodal multiscale attention block to perform subword alignment without the need of explicit alignment between language and acoustic cues. Finally, to empirically study the importance of language in this task, we extend the dataset proposed in Ahuja et al. (2020) with automatically extracted transcripts for audio signals. We substantiate the effectiveness of our approach through large-scale quantitative and user studies, which show that our proposed methodology significantly outperforms previous state-of-the-art approaches for gesture generation. Link to code, data and videos: https://github.com/chahuja/aisle

Automatic speech recognition systems usually require large annotated speech corpus for training. The manual annotation of a large corpus is very difficult. It can be very helpful to use unsupervised and semi-supervised learning methods in addition to supervised learning. In this work, we focus on using a semi-supervised training approach for Bangla Speech Recognition that can exploit large unpaired audio and text data. We encode speech and text data in an intermediate domain and propose a novel loss function based on the global encoding distance between encoded data to guide the semi-supervised training. Our proposed method reduces the Word Error Rate (WER) of the system from 37% to 31.9%.

Text summarization aims to extract essential information from a piece of text and transform the text into a concise version. Existing unsupervised abstractive summarization models leverage recurrent neural networks framework while the recently proposed transformer exhibits much more capability. Moreover, most of previous summarization models ignore abundant unlabeled corpora resources available for pretraining. In order to address these issues, we propose TED, a transformer-based unsupervised abstractive summarization system with pretraining on large-scale data. We first leverage the lead bias in news articles to pretrain the model on millions of unlabeled corpora. Next, we finetune TED on target domains through theme modeling and a denoising autoencoder to enhance the quality of generated summaries. Notably, TED outperforms all unsupervised abstractive baselines on NYT, CNN/DM and English Gigaword datasets with various document styles. Further analysis shows that the summaries generated by TED are highly abstractive, and each component in the objective function of TED is highly effective.