ACL 2022 Papers

Despite recent success, large neural models often generate factually incorrect text. Compounding this is the lack of a standard automatic evaluation for factuality–it cannot be meaningfully improved if it cannot be measured. Grounded generation promises a path to solving both of these problems: models draw on a reliable external document (grounding) for factual information, simplifying the challenge of factuality. Measuring factuality is also simplified–to factual consistency, testing whether the generation agrees with the grounding, rather than all facts. Yet, without a standard automatic metric for factual consistency, factually grounded generation remains an open problem. We study this problem for content transfer, in which generations extend a prompt, using information from factual grounding. Particularly, this domain allows us to introduce the notion of factual ablation for automatically measuring factual consistency: this captures the intuition that the model should be less likely to produce an output given a less relevant grounding document. In practice, we measure this by presenting a model with two grounding documents, and the model should prefer to use the more factually relevant one. We contribute two evaluation sets to measure this. Applying our new evaluation, we propose multiple novel methods improving over strong baselines.

Sampling is a promising bottom-up method for exposing what generative models have learned about language, but it remains unclear how to generate representative samples from popular masked language models (MLMs) like BERT. The MLM objective yields a dependency network with no guarantee of consistent conditional distributions, posing a problem for naive approaches. Drawing from theories of iterated learning in cognitive science, we explore the use of serial reproduction chains to sample from BERT’s priors. In particular, we observe that a unique and consistent estimator of the ground-truth joint distribution is given by a Generative Stochastic Network (GSN) sampler, which randomly selects which token to mask and reconstruct on each step. We show that the lexical and syntactic statistics of sentences from GSN chains closely match the ground-truth corpus distribution and perform better than other methods in a large corpus of naturalness judgments. Our findings establish a firmer theoretical foundation for bottom-up probing and highlight richer deviations from human priors.

Knowledge bases (KBs) contain plenty of structured world and commonsense knowledge. As such, they often complement distributional text-based information and facilitate various downstream tasks. Since their manual construction is resource- and time-intensive, recent efforts have tried leveraging large pretrained language models (PLMs) to generate additional monolingual knowledge facts for KBs. However, such methods have not been attempted for building and enriching multilingual KBs. Besides wider application, such multilingual KBs can provide richer combined knowledge than monolingual (e.g., English) KBs. Knowledge expressed in different languages may be complementary and unequally distributed: this implies that the knowledge available in high-resource languages can be transferred to low-resource ones. To achieve this, it is crucial to represent multilingual knowledge in a shared/unified space. To this end, we propose a unified representation model, Prix-LM, for multilingual KB construction and completion. We leverage two types of knowledge, monolingual triples and cross-lingual links, extracted from existing multilingual KBs, and tune a multilingual language encoder XLM-R via a causal language modeling objective. Prix-LM integrates useful multilingual and KB-based factual knowledge into a single model. Experiments on standard entity-related tasks, such as link prediction in multiple languages, cross-lingual entity linking and bilingual lexicon induction, demonstrate its effectiveness, with gains reported over strong task-specialised baselines.

This paper attacks the challenging problem of sign language translation (SLT), which involves not only visual and textual understanding but also additional prior knowledge learning (i.e. performing style, syntax). However, the majority of existing methods with vanilla encoder-decoder structures fail to sufficiently explore all of them. Based on this concern, we propose a novel method called Prior knowledge and memory Enriched Transformer (PET) for SLT, which incorporates the auxiliary information into vanilla transformer. Concretely, we develop gated interactive multi-head attention which associates the multimodal representation and global signing style with adaptive gated functions. One Part-of-Speech (POS) sequence generator relies on the associated information to predict the global syntactic structure, which is thereafter leveraged to guide the sentence generation. Besides, considering that the visual-textual context information, and additional auxiliary knowledge of a word may appear in more than one video, we design a multi-stream memory structure to obtain higher-quality translations, which stores the detailed correspondence between a word and its various relevant information, leading to a more comprehensive understanding for each word. We conduct extensive empirical studies on RWTH-PHOENIX-Weather-2014 dataset with both signer-dependent and signer-independent conditions. The quantitative and qualitative experimental results comprehensively reveal the effectiveness of PET.

Round-trip Machine Translation (MT) is a popular choice for paraphrase generation, which leverages readily available parallel corpora for supervision. In this paper, we formalize the implicit similarity function induced by this approach, and show that it is susceptible to non-paraphrase pairs sharing a single ambiguous translation. Based on these insights, we design an alternative similarity metric that mitigates this issue by requiring the entire translation distribution to match, and implement a relaxation of it through the Information Bottleneck method. Our approach incorporates an adversarial term into MT training in order to learn representations that encode as much information about the reference translation as possible, while keeping as little information about the input as possible. Paraphrases can be generated by decoding back to the source from this representation, without having to generate pivot translations. In addition to being more principled and efficient than round-trip MT, our approach offers an adjustable parameter to control the fidelity-diversity trade-off, and obtains better results in our experiments.

In this paper, we challenge the ACL community to reckon with historical and ongoing colonialism by adopting a set of ethical obligations and best practices drawn from the Indigenous studies literature. While the vast majority of NLP research focuses on a very small number of very high resource languages (English, Chinese, etc), some work has begun to engage with Indigenous languages. No research involving Indigenous language data can be considered ethical without first acknowledging that Indigenous languages are not merely very low resource languages. The toxic legacy of colonialism permeates every aspect of interaction between Indigenous communities and outside researchers. To this end, we propose that the ACL draft and adopt an ethical framework for NLP researchers and computational linguists wishing to engage in research involving Indigenous languages.

Recent advances in Automatic Speech Recognition (ASR) have made it possible to reliably produce automatic transcripts of clinician-patient conversations. However, access to clinical datasets is heavily restricted due to patient privacy, thus slowing down normal research practices. We detail the development of a public access, high quality dataset comprising of 57 mocked primary care consultations, including audio recordings, their manual utterance-level transcriptions, and the associated consultation notes. Our work illustrates how the dataset can be used as a benchmark for conversational medical ASR as well as consultation note generation from transcripts.

We investigate what kind of structural knowledge learned in neural network encoders is transferable to processing natural language. We design artificial languages with structural properties that mimic natural language, pretrain encoders on the data, and see how much performance the encoder exhibits on downstream tasks in natural language.Our experimental results show that pretraining with an artificial language with a nesting dependency structure provides some knowledge transferable to natural language.A follow-up probing analysis indicates that its success in the transfer is related to the amount of encoded contextual information and what is transferred is the knowledge of position-aware context dependence of language.Our results provide insights into how neural network encoders process human languages and the source of cross-lingual transferability of recent multilingual language models.

It is a common practice for recent works in vision language cross-modal reasoning to adopt a binary or multi-choice classification formulation taking as input a set of source image(s) and textual query. In this work, we take a sober look at such an “unconditional” formulation in the sense that no prior knowledge is specified with respect to the source image(s). Inspired by the designs of both visual commonsense reasoning and natural language inference tasks, we propose a new task termed “Premise-based Multi-modal Reasoning” (PMR) where a textual premise is the background presumption on each source image. The PMR dataset contains 15,360 manually annotated samples which are created by a multi-phase crowd-sourcing process. With selected high-quality movie screenshots and human-curated premise templates from 6 pre-defined categories, we ask crowd-source workers to write one true hypothesis and three distractors (4 choices) given the premise and image through a cross-check procedure.

Regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for NMT tasks in recent years. Despite their simplicity and effectiveness, we argue that these methods are limited by the under-fitting of training data. In this paper, we utilize prediction difference for ground-truth tokens to analyze the fitting of token-level samples and find that under-fitting is almost as common as over-fitting. We introduce prediction difference regularization (PD-R), a simple and effective method that can reduce over-fitting and under-fitting at the same time. For all token-level samples, PD-R minimizes the prediction difference between the original pass and the input-perturbed pass, making the model less sensitive to small input changes, thus more robust to both perturbations and under-fitted training data. Experiments on three widely used WMT translation tasks show that our approach can significantly improve over existing perturbation regularization methods. On WMT16 En-De task, our model achieves 1.80 SacreBLEU improvement over vanilla transformer.

Document-level text simplification often deletes some sentences besides performing lexical, grammatical or structural simplification to reduce text complexity. In this work, we focus on sentence deletions for text simplification and use a news genre-specific functional discourse structure, which categorizes sentences based on their contents and their function roles in telling a news story, for predicting sentence deletion. We incorporate sentence categories into a neural net model in two ways for predicting sentence deletions, either as additional features or by jointly predicting sentence deletions and sentence categories. Experimental results using human-annotated data show that incorporating the functional structure improves the recall of sentence deletion prediction by 6.5% and 10.7% respectively using the two methods, and improves the overall F1-score by 3.6% and 4.3% respectively.

Clinical trials offer a fundamental opportunity to discover new treatments and advance the medical knowledge. However, the uncertainty of the outcome of a trial can lead to unforeseen costs and setbacks. In this study, we propose a new method to predict the effectiveness of an intervention in a clinical trial. Our method relies on generating an informative summary from multiple documents available in the literature about the intervention under study. Specifically, our method first gathers all the abstracts of PubMed articles related to the intervention. Then, an evidence sentence, which conveys information about the effectiveness of the intervention, is extracted automatically from each abstract. Based on the set of evidence sentences extracted from the abstracts, a short summary about the intervention is constructed. Finally, the produced summaries are used to train a BERT-based classifier, in order to infer the effectiveness of an intervention. To evaluate our proposed method, we introduce a new dataset which is a collection of clinical trials together with their associated PubMed articles. Our experiments, demonstrate the effectiveness of producing short informative summaries and using them to predict the effectiveness of an intervention.

Item Response Theory (IRT) has been extensively used to numerically characterize question difficulty and discrimination for human subjects in domains including cognitive psychology and education (Primi et al., 2014; Downing, 2003). More recently, IRT has been used to similarly characterize item difficulty and discrimination for natural language models across various datasets (Lalor et al., 2019; Vania et al., 2021; Rodriguez et al., 2021). In this work, we explore predictive models for directly estimating and explaining these traits for natural language questions in a question-answering context. We use HotpotQA for illustration. Our experiments show that it is possible to predict both difficulty and discrimination parameters for new questions, and these traits are correlated with features of questions, answers, and associated contexts. Our findings can have significant implications for the creation of new datasets and tests on the one hand and strategies such as active learning and curriculum learning on the other.

Paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings. While cross-encoders have achieved high performances across several benchmarks, bi-encoders such as SBERT have been widely applied to sentence pair tasks. They exhibit substantially lower computation complexity and are better suited to symmetric tasks. In this work, we adopt a bi-encoder approach to the paraphrase identification task, and investigate the impact of explicitly incorporating predicate-argument information into SBERT through weighted aggregation. Experiments on six paraphrase identification datasets demonstrate that, with a minimal increase in parameters, the proposed model is able to outperform SBERT/SRoBERTa significantly. Further, ablation studies reveal that the predicate-argument based component plays a significant role in the performance gain.

Low-shot relation extraction (RE) aims to recognize novel relations with very few or even no samples, which is critical in real scenario application. Few-shot and zero-shot RE are two representative low-shot RE tasks, which seem to be with similar target but require totally different underlying abilities. In this paper, we propose Multi-Choice Matching Networks to unify low-shot relation extraction. To fill in the gap between zero-shot and few-shot RE, we propose the triplet-paraphrase meta-training, which leverages triplet paraphrase to pre-train zero-shot label matching ability and uses meta-learning paradigm to learn few-shot instance summarizing ability. Experimental results on three different low-shot RE tasks show that the proposed method outperforms strong baselines by a large margin, and achieve the best performance on few-shot RE leaderboard.

Recent years have witnessed growing interests in incorporating external knowledge such as pre-trained word embeddings (PWEs) or pre-trained language models (PLMs) into neural topic modeling. However, we found that employing PWEs and PLMs for topic modeling only achieved limited performance improvements but with huge computational overhead. In this paper, we propose a novel strategy to incorporate external knowledge into neural topic modeling where the neural topic model is pre-trained on a large corpus and then fine-tuned on the target dataset. Experiments have been conducted on three datasets and results show that the proposed approach significantly outperforms both current state-of-the-art neural topic models and some topic modeling approaches enhanced with PWEs or PLMs. Moreover, further study shows that the proposed approach greatly reduces the need for the huge size of training data.

What can pre-trained multilingual sequence-to-sequence models like mBART contribute to translating low-resource languages? We conduct a thorough empirical experiment in 10 languages to ascertain this, considering five factors: (1) the amount of fine-tuning data, (2) the noise in the fine-tuning data, (3) the amount of pre-training data in the model, (4) the impact of domain mismatch, and (5) language typology. In addition to yielding several heuristics, the experiments form a framework for evaluating the data sensitivities of machine translation systems. While mBART is robust to domain differences, its translations for unseen and typologically distant languages remain below 3.0 BLEU. In answer to our title’s question, mBART is not a low-resource panacea; we therefore encourage shifting the emphasis from new models to new data.

Language models (LMs) have shown great potential as implicit knowledge bases (KBs). And for their practical use, knowledge in LMs need to be updated periodically. However, existing tasks to assess LMs’ efficacy as KBs do not adequately consider multiple large-scale updates. To this end, we first propose a novel task—Continuously-updated QA (CuQA)—in which multiple large-scale updates are made to LMs, and the performance is measured with respect to the success in adding and updating knowledge while retaining existing knowledge. We then present LMs with plug-in modules that effectively handle the updates. Experiments conducted on zsRE QA and NQ datasets show that our method outperforms existing approaches. We find that our method is 4x more effective in terms of updates/forgets ratio, compared to a fine-tuning baseline.

Modern NLP classifiers are known to return uncalibrated estimations of class posteriors. Existing methods for posterior calibration rescale the predicted probabilities but often have an adverse impact on final classification accuracy, thus leading to poorer generalization. We propose an end-to-end trained calibrator, Platt-Binning, that directly optimizes the objective while minimizing the difference between the predicted and empirical posterior probabilities. Our method leverages the sample efficiency of Platt scaling and the verification guarantees of histogram binning, thus not only reducing the calibration error but also improving task performance. In contrast to existing calibrators, we perform this efficient calibration during training. Empirical evaluation of benchmark NLP classification tasks echoes the efficacy of our proposal.