ACL 2020 Papers

Context gates are effective to control the contributions from the source and target contexts in the recurrent neural network (RNN) based neural machine translation (NMT). However, it is challenging to extend them into the advanced Transformer architecture, which is more complicated than RNN. This paper first provides a method to identify source and target contexts and then introduce a gate mechanism to control the source and target contributions in Transformer. In addition, to further reduce the bias problem in the gate mechanism, this paper proposes a regularization method to guide the learning of the gates with supervision automatically generated using pointwise mutual information. Extensive experiments on 4 translation datasets demonstrate that the proposed model obtains an averaged gain of 1.0 BLEU score over a strong Transformer baseline.

Web-crawled data provides a good source of parallel corpora for training machine translation models. It is automatically obtained, but extremely noisy, and recent work shows that neural machine translation systems are more sensitive to noise than traditional statistical machine translation methods. In this paper, we propose a novel approach to filter out noisy sentence pairs from web-crawled corpora via pre-trained language models. We measure sentence parallelism by leveraging the multilingual capability of BERT and use the Generative Pre-training (GPT) language model as a domain filter to balance data domains. We evaluate the proposed method on the WMT 2018 Parallel Corpus Filtering shared task, and on our own web-crawled Japanese-Chinese parallel corpus. Our method significantly outperforms baselines and achieves a new state-of-the-art. In an unsupervised setting, our method achieves comparable performance to the top-1 supervised method. We also evaluate on a web-crawled Japanese-Chinese parallel corpus that we make publicly available.

Neural Machine Translation (NMT) models are sensitive to small perturbations in the input. Robustness to such perturbations is typically measured using translation quality metrics such as BLEU on the noisy input. This paper proposes additional metrics which measure the relative degradation and changes in translation when small perturbations are added to the input. We focus on a class of models employing subword regularization to address robustness and perform extensive evaluations of these models using the robustness measures proposed. Results show that our proposed metrics reveal a clear trend of improved robustness to perturbations when subword regularization methods are used.

When training multilingual machine translation (MT) models that can translate to/from multiple languages, we are faced with imbalanced training sets: some languages have much more training data than others. Standard practice is to up-sample less resourced languages to increase representation, and the degree of up-sampling has a large effect on the overall performance. In this paper, we propose a method that instead automatically learns how to weight training data through a data scorer that is optimized to maximize performance on all test languages. Experiments on two sets of languages under both one-to-many and many-to-one MT settings show our method not only consistently outperforms heuristic baselines in terms of average performance, but also offers flexible control over the performance of which languages are optimized.

This paper proposes a simple and effective approach to address the problem of posterior collapse in conditional variational autoencoders (CVAEs). It thus improves performance of machine translation models that use noisy or monolingual data, as well as in conventional settings. Extending Transformer and conditional VAEs, our proposed latent variable model measurably prevents posterior collapse by (1) using a modified evidence lower bound (ELBO) objective which promotes mutual information between the latent variable and the target, and (2) guiding the latent variable with an auxiliary bag-of-words prediction task. As a result, the proposed model yields improved translation quality compared to existing variational NMT models on WMT Ro↔En and De↔En. With latent variables being effectively utilized, our model demonstrates improved robustness over non-latent Transformer in handling uncertainty: exploiting noisy source-side monolingual data (up to +3.2 BLEU), and training with weakly aligned web-mined parallel data (up to +4.7 BLEU).

Training neural models for named entity recognition (NER) in a new domain often requires additional human annotations (e.g., tens of thousands of labeled instances) that are usually expensive and time-consuming to collect. Thus, a crucial research question is how to obtain supervision in a cost-effective way. In this paper, we introduce “entity triggers,” an effective proxy of human explanations for facilitating label-efficient learning of NER models. An entity trigger is defined as a group of words in a sentence that helps to explain why humans would recognize an entity in the sentence. We crowd-sourced 14k entity triggers for two well-studied NER datasets. Our proposed model, Trigger Matching Network, jointly learns trigger representations and soft matching module with self-attention such that can generalize to unseen sentences easily for tagging. Our framework is significantly more cost-effective than the traditional neural NER frameworks. Experiments show that using only 20% of the trigger-annotated sentences results in a comparable performance as using 70% of conventional annotated sentences.

Extracting structured knowledge from product profiles is crucial for various applications in e-Commerce. State-of-the-art approaches for knowledge extraction were each designed for a single category of product, and thus do not apply to real-life e-Commerce scenarios, which often contain thousands of diverse categories. This paper proposes TXtract, a taxonomy-aware knowledge extraction model that applies to thousands of product categories organized in a hierarchical taxonomy. Through category conditional self-attention and multi-task learning, our approach is both scalable, as it trains a single model for thousands of categories, and effective, as it extracts category-specific attribute values. Experiments on products from a taxonomy with 4,000 categories show that TXtract outperforms state-of-the-art approaches by up to 10% in F1 and 15% in coverage across all categories.

Named entity recognition is a key component of many text processing pipelines and it is thus essential for this component to be robust to different types of input. However, domain transfer of NER models with data from multiple genres has not been widely studied. To this end, we conduct NER experiments in three predictive setups on data from: a) multiple domains; b) multiple domains where the genre label is unknown at inference time; c) domains not encountered in training. We introduce a new architecture tailored to this task by using shared and private domain parameters and multi-task learning. This consistently outperforms all other baseline and competitive methods on all three experimental setups, with differences ranging between +1.95 to +3.11 average F1 across multiple genres when compared to standard approaches. These results illustrate the challenges that need to be taken into account when building real-world NLP applications that are robust to various types of text and the methods that can help, at least partially, alleviate these issues.

We propose a novel method for hierarchical entity classification that embraces ontological structure at both training and during prediction. At training, our novel multi-level learning-to-rank loss compares positive types against negative siblings according to the type tree. During prediction, we define a coarse-to-fine decoder that restricts viable candidates at each level of the ontology based on already predicted parent type(s). Our approach significantly outperform prior work on strict accuracy, demonstrating the effectiveness of our method.

Concept normalization, the task of linking textual mentions of concepts to concepts in an ontology, is challenging because ontologies are large. In most cases, annotated datasets cover only a small sample of the concepts, yet concept normalizers are expected to predict all concepts in the ontology. In this paper, we propose an architecture consisting of a candidate generator and a list-wise ranker based on BERT. The ranker considers pairings of concept mentions and candidate concepts, allowing it to make predictions for any concept, not just those seen during training. We further enhance this list-wise approach with a semantic type regularizer that allows the model to incorporate semantic type information from the ontology during training. Our proposed concept normalization framework achieves state-of-the-art performance on multiple datasets.

This paper shows that pretraining multilingual language models at scale leads to significant performance gains for a wide range of cross-lingual transfer tasks. We train a Transformer-based masked language model on one hundred languages, using more than two terabytes of filtered CommonCrawl data. Our model, dubbed XLM-R, significantly outperforms multilingual BERT (mBERT) on a variety of cross-lingual benchmarks, including +14.6% average accuracy on XNLI, +13% average F1 score on MLQA, and +2.4% F1 score on NER. XLM-R performs particularly well on low-resource languages, improving 15.7% in XNLI accuracy for Swahili and 11.4% for Urdu over previous XLM models. We also present a detailed empirical analysis of the key factors that are required to achieve these gains, including the trade-offs between (1) positive transfer and capacity dilution and (2) the performance of high and low resource languages at scale. Finally, we show, for the first time, the possibility of multilingual modeling without sacrificing per-language performance; XLM-R is very competitive with strong monolingual models on the GLUE and XNLI benchmarks. We will make our code and models publicly available.

We introduce a transductive model for parsing into Universal Decompositional Semantics (UDS) representations, which jointly learns to map natural language utterances into UDS graph structures and annotate the graph with decompositional semantic attribute scores. We also introduce a strong pipeline model for parsing into the UDS graph structure, and show that our transductive parser performs comparably while additionally performing attribute prediction. By analyzing the attribute prediction errors, we find the model captures natural relationships between attribute groups.

Recent years have witnessed the burgeoning of pretrained language models (LMs) for text-based natural language (NL) understanding tasks. Such models are typically trained on free-form NL text, hence may not be suitable for tasks like semantic parsing over structured data, which require reasoning over both free-form NL questions and structured tabular data (e.g., database tables). In this paper we present TaBERT, a pretrained LM that jointly learns representations for NL sentences and (semi-)structured tables. TaBERT is trained on a large corpus of 26 million tables and their English contexts. In experiments, neural semantic parsers using TaBERT as feature representation layers achieve new best results on the challenging weakly-supervised semantic parsing benchmark WikiTableQuestions, while performing competitively on the text-to-SQL dataset Spider.

Recent neural network-driven semantic role labeling (SRL) systems have shown impressive improvements in F1 scores. These improvements are due to expressive input representations, which, at least at the surface, are orthogonal to knowledge-rich constrained decoding mechanisms that helped linear SRL models. Introducing the benefits of structure to inform neural models presents a methodological challenge. In this paper, we present a structured tuning framework to improve models using softened constraints only at training time. Our framework leverages the expressiveness of neural networks and provides supervision with structured loss components. We start with a strong baseline (RoBERTa) to validate the impact of our approach, and show that our framework outperforms the baseline by learning to comply with declarative constraints. Additionally, our experiments with smaller training sizes show that we can achieve consistent improvements under low-resource scenarios.

The focus of a negation is the set of tokens intended to be negated, and a key component for revealing affirmative alternatives to negated utterances. In this paper, we experiment with neural networks to predict the focus of negation. Our main novelty is leveraging a scope detector to introduce the scope of negation as an additional input to the network. Experimental results show that doing so obtains the best results to date. Additionally, we perform a detailed error analysis providing insights into the main error categories, and analyze errors depending on whether the model takes into account scope and context information.

We study the task of cross-database semantic parsing (XSP), where a system that maps natural language utterances to executable SQL queries is evaluated on databases unseen during training. Recently, several datasets, including Spider, were proposed to support development of XSP systems. We propose a challenging evaluation setup for cross-database semantic parsing, focusing on variation across database schemas and in-domain language use. We re-purpose eight semantic parsing datasets that have been well-studied in the setting where in-domain training data is available, and instead use them as additional evaluation data for XSP systems instead. We build a system that performs well on Spider, and find that it struggles to generalize to our re-purposed set. Our setup uncovers several generalization challenges for cross-database semantic parsing, demonstrating the need to use and develop diverse training and evaluation datasets.

Word embedding-based similarity measures are currently among the top-performing methods on unsupervised semantic textual similarity (STS) tasks. Recent work has increasingly adopted a statistical view on these embeddings, with some of the top approaches being essentially various correlations (which include the famous cosine similarity). Another excellent candidate for a similarity measure is mutual information (MI), which can capture arbitrary dependencies between the variables and has a simple and intuitive expression. Unfortunately, its use in the context of dense word embeddings has so far been avoided due to difficulties with estimating MI for continuous data. In this work we go through a vast literature on estimating MI in such cases and single out the most promising methods, yielding a simple and elegant similarity measure for word embeddings. We show that mutual information is a viable alternative to correlations, gives an excellent signal that correlates well with human judgements of similarity and rivals existing state-of-the-art unsupervised methods.

Language models pretrained on text from a wide variety of sources form the foundation of today’s NLP. In light of the success of these broad-coverage models, we investigate whether it is still helpful to tailor a pretrained model to the domain of a target task. We present a study across four domains (biomedical and computer science publications, news, and reviews) and eight classification tasks, showing that a second phase of pretraining in-domain (domain-adaptive pretraining) leads to performance gains, under both high- and low-resource settings. Moreover, adapting to the task’s unlabeled data (task-adaptive pretraining) improves performance even after domain-adaptive pretraining. Finally, we show that adapting to a task corpus augmented using simple data selection strategies is an effective alternative, especially when resources for domain-adaptive pretraining might be unavailable. Overall, we consistently find that multi-phase adaptive pretraining offers large gains in task performance.

This paper introduces two tasks: determining (a) the duration of possession relations and (b) co-possessions, i.e., whether multiple possessors possess a possessee at the same time. We present new annotations on top of corpora annotating possession existence and experimental results. Regarding possession duration, we derive the time spans we work with empirically from annotations indicating lower and upper bounds. Regarding co-possessions, we use a binary label. Cohen’s kappa coefficients indicate substantial agreement, and experimental results show that text is more useful than the image for solving these tasks.