NAACL 2021 Papers

Natural language inference requires reasoning about contradictions, negations, and their commonsense implications. Given a simple premise (e.g., “I’m mad at you”), humans can reason about the varying shades of contradictory statements ranging from straightforward negations (“I’m not mad at you”) to commonsense contradictions (“I’m happy”). Moreover, these negated or contradictory statements shift the commonsense implications of the original premise in interesting and nontrivial ways. For example, while “I’m mad” implies “I’m unhappy about something,” negating the premise does not necessarily negate the corresponding commonsense implications. In this paper, we present the first comprehensive study focusing on commonsense implications of negated statements and contradictions. We introduce ANION, a new commonsense knowledge graph with 624K if-then rules focusing on negated and contradictory events. We then present joint generative and discriminative inference models for this new resource, providing novel empirical insights on how logical negations and commonsense contradictions reshape the commonsense implications of their original premises.

The stance detection task aims at detecting the stance of a tweet or a text for a target. These targets can be named entities or free-form sentences (claims). Though the task involves reasoning of the tweet with respect to a target, we find that it is possible to achieve high accuracy on several publicly available Twitter stance detection datasets without looking at the target sentence. Specifically, a simple tweet classification model achieved human-level performance on the WT–WT dataset and more than two-third accuracy on various other datasets. We investigate the existence of biases in such datasets to find the potential spurious correlations of sentiment-stance relations and lexical choice associated with the stance category. Furthermore, we propose a new large dataset free of such biases and demonstrate its aptness on the existing stance detection systems. Our empirical findings show much scope for research on the stance detection task and proposes several considerations for creating future stance detection datasets.

We focus on a type of linguistic formal reasoning where the goal is to reason over explicit knowledge in the form of natural language facts and rules (Clark et al., 2020). A recent work, named PRover (Saha et al., 2020), performs such reasoning by answering a question and also generating a proof graph that explains the answer. However, compositional reasoning is not always unique and there may be multiple ways of reaching the correct answer. Thus, in our work, we address a new and challenging problem of generating multiple proof graphs for reasoning over natural language rule-bases. Each proof provides a different rationale for the answer, thereby improving the interpretability of such reasoning systems. In order to jointly learn from all proof graphs and exploit the correlations between multiple proofs for a question, we pose this task as a set generation problem over structured output spaces where each proof is represented as a directed graph. We propose two variants of a proof-set generation model, multiPRover. Our first model, Multilabel-multiPRover, generates a set of proofs via multi-label classification and implicit conditioning between the proofs; while the second model, Iterative-multiPRover, generates proofs iteratively by explicitly conditioning on the previously generated proofs. Experiments on multiple synthetic, zero-shot, and human-paraphrased datasets reveal that both multiPRover models significantly outperform PRover on datasets containing multiple gold proofs. Iterative-multiPRover obtains state-of-the-art proof F1 in zero-shot scenarios where all examples have single correct proofs. It also generalizes better to questions requiring higher depths of reasoning where multiple proofs are more frequent.

The recent “Text-to-Text Transfer Transformer” (T5) leveraged a unified text-to-text format and scale to attain state-of-the-art results on a wide variety of English-language NLP tasks. In this paper, we introduce mT5, a multilingual variant of T5 that was pre-trained on a new Common Crawl-based dataset covering 101 languages. We detail the design and modified training of mT5 and demonstrate its state-of-the-art performance on many multilingual benchmarks. We also describe a simple technique to prevent “accidental translation” in the zero-shot setting, where a generative model chooses to (partially) translate its prediction into the wrong language. All of the code and model checkpoints used in this work are publicly available.

While relation extraction is an essential task in knowledge acquisition and representation, and new-generated relations are common in the real world, less effort is made to predict unseen relations that cannot be observed at the training stage. In this paper, we formulate the zero-shot relation extraction problem by incorporating the text description of seen and unseen relations. We propose a novel multi-task learning model, Zero-Shot BERT (ZS-BERT), to directly predict unseen relations without hand-crafted attribute labeling and multiple pairwise classifications. Given training instances consisting of input sentences and the descriptions of their seen relations, ZS-BERT learns two functions that project sentences and relations into an embedding space by jointly minimizing the distances between them and classifying seen relations. By generating the embeddings of unseen relations and new-coming sentences based on such two functions, we use nearest neighbor search to obtain the prediction of unseen relations. Experiments conducted on two well-known datasets exhibit that ZS-BERT can outperform existing methods by at least 13.54% improvement on F1 score.

The increasing popularity of voice-based personal assistants provides new opportunities for conversational recommendation. One particularly interesting area is movie recommendation, which can benefit from an open-ended interaction with the user, through a natural conversation. We explore one promising direction for conversational recommendation: mapping a conversational user, for whom there is limited or no data available, to most similar external reviewers, whose preferences are known, by representing the conversation as a user’s interest vector, and adapting collaborative filtering techniques to estimate the current user’s preferences for new movies. We call our proposed method ConvExtr (Conversational Collaborative Filtering using External Data), which 1) infers a user’s sentiment towards an entity from the conversation context, and 2) transforms the ratings of “similar” external reviewers to predict the current user’s preferences. We implement these steps by adapting contextual sentiment prediction techniques, and domain adaptation, respectively. To evaluate our method, we develop and make available a finely annotated dataset of movie recommendation conversations, which we call MovieSent. Our results demonstrate that ConvExtr can improve the accuracy of predicting users’ ratings for new movies by exploiting conversation content and external data.

Multilingual question answering tasks typically assume that answers exist in the same language as the question. Yet in practice, many languages face both information scarcity—where languages have few reference articles—and information asymmetry—where questions reference concepts from other cultures. This work extends open-retrieval question answering to a cross-lingual setting enabling questions from one language to be answered via answer content from another language. We construct a large-scale dataset built on 40K information-seeking questions across 7 diverse non-English languages that TyDi QA could not find same-language answers for. Based on this dataset, we introduce a task framework, called Cross-lingual Open-Retrieval Question Answering (XOR QA), that consists of three new tasks involving cross-lingual document retrieval from multilingual and English resources. We establish baselines with state-of-the-art machine translation systems and cross-lingual pretrained models. Experimental results suggest that XOR QA is a challenging task that will facilitate the development of novel techniques for multilingual question answering. Our data and code are available at https://nlp.cs.washington.edu/xorqa/.

Multilingual models, such as M-BERT and XLM-R, have gained increasing popularity, due to their zero-shot cross-lingual transfer learning capabilities. However, their generalization ability is still inconsistent for typologically diverse languages and across different benchmarks. Recently, meta-learning has garnered attention as a promising technique for enhancing transfer learning under low-resource scenarios: particularly for cross-lingual transfer in Natural Language Understanding (NLU). In this work, we propose X-METRA-ADA, a cross-lingual MEta-TRAnsfer learning ADAptation approach for NLU. Our approach adapts MAML, an optimization-based meta-learning approach, to learn to adapt to new languages. We extensively evaluate our framework on two challenging cross-lingual NLU tasks: multilingual task-oriented dialog and typologically diverse question answering. We show that our approach outperforms naive fine-tuning, reaching competitive performance on both tasks for most languages. Our analysis reveals that X-METRA-ADA can leverage limited data for faster adaptation.

In this paper, we explore text classification with extremely weak supervision, i.e., only relying on the surface text of class names. This is a more challenging setting than the seed-driven weak supervision, which allows a few seed words per class. We opt to attack this problem from a representation learning perspective—ideal document representations should lead to nearly the same results between clustering and the desired classification. In particular, one can classify the same corpus differently (e.g., based on topics and locations), so document representations should be adaptive to the given class names. We propose a novel framework X-Class to realize the adaptive representations. Specifically, we first estimate class representations by incrementally adding the most similar word to each class until inconsistency arises. Following a tailored mixture of class attention mechanisms, we obtain the document representation via a weighted average of contextualized word representations. With the prior of each document assigned to its nearest class, we then cluster and align the documents to classes. Finally, we pick the most confident documents from each cluster to train a text classifier. Extensive experiments demonstrate that X-Class can rival and even outperform seed-driven weakly supervised methods on 7 benchmark datasets.

Although Question-Answering has long been of research interest, its accessibility to users through a speech interface and its support to multiple languages have not been addressed in prior studies. Towards these ends, we present a new task and a synthetically-generated dataset to do Fact-based Visual Spoken-Question Answering (FVSQA). FVSQA is based on the FVQA dataset, which requires a system to retrieve an entity from Knowledge Graphs (KGs) to answer a question about an image. In FVSQA, the question is spoken rather than typed. Three sub-tasks are proposed: (1) speech-to-text based, (2) end-to-end, without speech-to-text as an intermediate component, and (3) cross-lingual, in which the question is spoken in a language different from that in which the KG is recorded. The end-to-end and cross-lingual tasks are the first to require world knowledge from a multi-relational KG as a differentiable layer in an end-to-end spoken language understanding task, hence the proposed reference implementation is called Worldly-Wise (WoW).WoW is shown to perform end-to-end cross-lingual FVSQA at same levels of accuracy across 3 languages - English, Hindi, and Turkish.

Lexical complexity is a highly subjective notion, yet this factor is often neglected in lexical simplification and readability systems which use a ”one-size-fits-all” approach. In this paper, we investigate which aspects contribute to the notion of lexical complexity in various groups of readers, focusing on native and non-native speakers of English, and how the notion of complexity changes depending on the proficiency level of a non-native reader. To facilitate reproducibility of our approach and foster further research into these aspects, we release a dataset of complex words annotated by readers with different backgrounds.

Masked language models have quickly become the de facto standard when processing text. Recently, several approaches have been proposed to further enrich word representations with external knowledge sources such as knowledge graphs. However, these models are devised and evaluated in a monolingual setting only. In this work, we propose a language-independent entity prediction task as an intermediate training procedure to ground word representations on entity semantics and bridge the gap across different languages by means of a shared vocabulary of entities. We show that our approach effectively injects new lexical-semantic knowledge into neural models, improving their performance on different semantic tasks in the zero-shot crosslingual setting. As an additional advantage, our intermediate training does not require any supplementary input, allowing our models to be applied to new datasets right away. In our experiments, we use Wikipedia articles in up to 100 languages and already observe consistent gains compared to strong baselines when predicting entities using only the English Wikipedia. Further adding extra languages lead to improvements in most tasks up to a certain point, but overall we found it non-trivial to scale improvements in model transferability by training on ever increasing amounts of Wikipedia languages.

This study introduces and analyzes WikiTalkEdit, a dataset of conversations and edit histories from Wikipedia, for research in online cooperation and conversation modeling. The dataset comprises dialog triplets from the Wikipedia Talk pages, and editing actions on the corresponding articles being discussed. We show how the data supports the classic understanding of style matching, where positive emotion and the use of first-person pronouns predict a positive emotional change in a Wikipedia contributor. However, they do not predict editorial behavior. On the other hand, feedback invoking evidentiality and criticism, and references to Wikipedia’s community norms, is more likely to persuade the contributor to perform edits but is less likely to lead to a positive emotion. We developed baseline classifiers trained on pre-trained RoBERTa features that can predict editorial change with an F1 score of .54, as compared to an F1 score of .66 for predicting emotional change. A diagnostic analysis of persisting errors is also provided. We conclude with possible applications and recommendations for future work. The dataset is publicly available for the research community at https://github.com/kj2013/WikiTalkEdit/.

Machine learning solutions are often criticized for the lack of explanation of their successes and failures. Understanding which instances are misclassified and why is essential to improve the learning process. This work helps to fill this gap by proposing a methodology to characterize, quantify and measure the impact of hard instances in the task of polarity classification of movie reviews. We characterize such instances into two categories: neutrality, where the text does not convey a clear polarity, and discrepancy, where the polarity of the text is the opposite of its true rating. We quantify the number of hard instances in polarity classification of movie reviews and provide empirical evidence about the need to pay attention to such problematic instances, as they are much harder to classify, for both machine and human classifiers. To the best of our knowledge, this is the first systematic analysis of the impact of hard instances in polarity detection from well-formed textual reviews.

Email and chat communication tools are increasingly important for completing daily tasks. Accurate real-time phrase completion can save time and bolster productivity. Modern text prediction algorithms are based on large language models which typically rely on the prior words in a message to predict a completion. We examine how additional contextual signals (from previous messages, time, and subject) affect the performance of a commercial text prediction model. We compare contextual text prediction in chat and email messages from two of the largest commercial platforms Microsoft Teams and Outlook, finding that contextual signals contribute to performance differently between these scenarios. On emails, time context is most beneficial with small relative gains of 2% over baseline. Whereas, in chat scenarios, using a tailored set of previous messages as context yields relative improvements over the baseline between 9.3% and 18.6% across various critical service-oriented text prediction metrics.

Although deep neural networks have been widely employed and proven effective in sentiment analysis tasks, it remains challenging for model developers to assess their models for erroneous predictions that might exist prior to deployment. Once deployed, emergent errors can be hard to identify in prediction run-time and impossible to trace back to their sources. To address such gaps, in this paper we propose an error detection framework for sentiment analysis based on explainable features. We perform global-level feature validation with human-in-the-loop assessment, followed by an integration of global and local-level feature contribution analysis. Experimental results show that, given limited human-in-the-loop intervention, our method is able to identify erroneous model predictions on unseen data with high precision.

Transfer learning based on pretraining language models on a large amount of raw data has become a new norm to reach state-of-the-art performance in NLP. Still, it remains unclear how this approach should be applied for unseen languages that are not covered by any available large-scale multilingual language model and for which only a small amount of raw data is generally available. In this work, by comparing multilingual and monolingual models, we show that such models behave in multiple ways on unseen languages. Some languages greatly benefit from transfer learning and behave similarly to closely related high resource languages whereas others apparently do not. Focusing on the latter, we show that this failure to transfer is largely related to the impact of the script used to write such languages. We show that transliterating those languages significantly improves the potential of large-scale multilingual language models on downstream tasks. This result provides a promising direction towards making these massively multilingual models useful for a new set of unseen languages.

Summarization of clinical narratives is a long-standing research problem. Here, we introduce the task of hospital-course summarization. Given the documentation authored throughout a patient’s hospitalization, generate a paragraph that tells the story of the patient admission. We construct an English, text-to-text dataset of 109,000 hospitalizations (2M source notes) and their corresponding summary proxy: the clinician-authored “Brief Hospital Course” paragraph written as part of a discharge note. Exploratory analyses reveal that the BHC paragraphs are highly abstractive with some long extracted fragments; are concise yet comprehensive; differ in style and content organization from the source notes; exhibit minimal lexical cohesion; and represent silver-standard references. Our analysis identifies multiple implications for modeling this complex, multi-document summarization task.

Evaluation for many natural language understanding (NLU) tasks is broken: Unreliable and biased systems score so highly on standard benchmarks that there is little room for researchers who develop better systems to demonstrate their improvements. The recent trend to abandon IID benchmarks in favor of adversarially-constructed, out-of-distribution test sets ensures that current models will perform poorly, but ultimately only obscures the abilities that we want our benchmarks to measure. In this position paper, we lay out four criteria that we argue NLU benchmarks should meet. We argue most current benchmarks fail at these criteria, and that adversarial data collection does not meaningfully address the causes of these failures. Instead, restoring a healthy evaluation ecosystem will require significant progress in the design of benchmark datasets, the reliability with which they are annotated, their size, and the ways they handle social bias.