ACL 2020 论文列表

We report on methods to create the largest publicly available parallel corpora by crawling the web, using open source software. We empirically compare alternative methods and publish benchmark data sets for sentence alignment and sentence pair filtering. We also describe the parallel corpora released and evaluate their quality and their usefulness to create machine translation systems.

Discourse representation structures (DRSs) are scoped semantic representations for texts of arbitrary length. Evaluating the accuracy of predicted DRSs plays a key role in developing semantic parsers and improving their performance. DRSs are typically visualized as boxes which are not straightforward to process automatically. Counter transforms DRSs to clauses and measures clause overlap by searching for variable mappings between two DRSs. However, this metric is computationally costly (with respect to memory and CPU time) and does not scale with longer texts. We introduce Dscorer, an efficient new metric which converts box-style DRSs to graphs and then measures the overlap of n-grams. Experiments show that Dscorer computes accuracy scores that are correlated with Counter at a fraction of the time.

A major hurdle in data-driven research on typology is having sufficient data in many languages to draw meaningful conclusions. We present VoxClamantis v1.0, the first large-scale corpus for phonetic typology, with aligned segments and estimated phoneme-level labels in 690 readings spanning 635 languages, along with acoustic-phonetic measures of vowels and sibilants. Access to such data can greatly facilitate investigation of phonetic typology at a large scale and across many languages. However, it is non-trivial and computationally intensive to obtain such alignments for hundreds of languages, many of which have few to no resources presently available. We describe the methodology to create our corpus, discuss caveats with current methods and their impact on the utility of this data, and illustrate possible research directions through a series of case studies on the 48 highest-quality readings. Our corpus and scripts are publicly available for non-commercial use at https://voxclamantisproject.github.io.

Evidence retrieval is a critical stage of question answering (QA), necessary not only to improve performance, but also to explain the decisions of the QA method. We introduce a simple, fast, and unsupervised iterative evidence retrieval method, which relies on three ideas: (a) an unsupervised alignment approach to soft-align questions and answers with justification sentences using only GloVe embeddings, (b) an iterative process that reformulates queries focusing on terms that are not covered by existing justifications, which (c) stops when the terms in the given question and candidate answers are covered by the retrieved justifications. Despite its simplicity, our approach outperforms all the previous methods (including supervised methods) on the evidence selection task on two datasets: MultiRC and QASC. When these evidence sentences are fed into a RoBERTa answer classification component, we achieve state-of-the-art QA performance on these two datasets.

Question Answering (QA) is in increasing demand as the amount of information available online and the desire for quick access to this content grows. A common approach to QA has been to fine-tune a pretrained language model on a task-specific labeled dataset. This paradigm, however, relies on scarce, and costly to obtain, large-scale human-labeled data. We propose an unsupervised approach to training QA models with generated pseudo-training data. We show that generating questions for QA training by applying a simple template on a related, retrieved sentence rather than the original context sentence improves downstream QA performance by allowing the model to learn more complex context-question relationships. Training a QA model on this data gives a relative improvement over a previous unsupervised model in F1 score on the SQuAD dataset by about 14%, and 20% when the answer is a named entity, achieving state-of-the-art performance on SQuAD for unsupervised QA.

Knowledge Graphs (KG) are multi-relational graphs consisting of entities as nodes and relations among them as typed edges. Goal of the Question Answering over KG (KGQA) task is to answer natural language queries posed over the KG. Multi-hop KGQA requires reasoning over multiple edges of the KG to arrive at the right answer. KGs are often incomplete with many missing links, posing additional challenges for KGQA, especially for multi-hop KGQA. Recent research on multi-hop KGQA has attempted to handle KG sparsity using relevant external text, which isn’t always readily available. In a separate line of research, KG embedding methods have been proposed to reduce KG sparsity by performing missing link prediction. Such KG embedding methods, even though highly relevant, have not been explored for multi-hop KGQA so far. We fill this gap in this paper and propose EmbedKGQA. EmbedKGQA is particularly effective in performing multi-hop KGQA over sparse KGs. EmbedKGQA also relaxes the requirement of answer selection from a pre-specified neighborhood, a sub-optimal constraint enforced by previous multi-hop KGQA methods. Through extensive experiments on multiple benchmark datasets, we demonstrate EmbedKGQA’s effectiveness over other state-of-the-art baselines.

Transformer-based QA models use input-wide self-attention – i.e. across both the question and the input passage – at all layers, causing them to be slow and memory-intensive. It turns out that we can get by without input-wide self-attention at all layers, especially in the lower layers. We introduce DeFormer, a decomposed transformer, which substitutes the full self-attention with question-wide and passage-wide self-attentions in the lower layers. This allows for question-independent processing of the input text representations, which in turn enables pre-computing passage representations reducing runtime compute drastically. Furthermore, because DeFormer is largely similar to the original model, we can initialize DeFormer with the pre-training weights of a standard transformer, and directly fine-tune on the target QA dataset. We show DeFormer versions of BERT and XLNet can be used to speed up QA by over 4.3x and with simple distillation-based losses they incur only a 1% drop in accuracy. We open source the code at https://github.com/StonyBrookNLP/deformer.

Machine reading comprehension has made great progress in recent years owing to large-scale annotated datasets. In the clinical domain, however, creating such datasets is quite difficult due to the domain expertise required for annotation. Recently, Pampari et al. (EMNLP’18) tackled this issue by using expert-annotated question templates and existing i2b2 annotations to create emrQA, the first large-scale dataset for question answering (QA) based on clinical notes. In this paper, we provide an in-depth analysis of this dataset and the clinical reading comprehension (CliniRC) task. From our qualitative analysis, we find that (i) emrQA answers are often incomplete, and (ii) emrQA questions are often answerable without using domain knowledge. From our quantitative experiments, surprising results include that (iii) using a small sampled subset (5%-20%), we can obtain roughly equal performance compared to the model trained on the entire dataset, (iv) this performance is close to human expert’s performance, and (v) BERT models do not beat the best performing base model. Following our analysis of the emrQA, we further explore two desired aspects of CliniRC systems: the ability to utilize clinical domain knowledge and to generalize to unseen questions and contexts. We argue that both should be considered when creating future datasets.

In many settings it is important for one to be able to understand why a model made a particular prediction. In NLP this often entails extracting snippets of an input text ‘responsible for’ corresponding model output; when such a snippet comprises tokens that indeed informed the model’s prediction, it is a faithful explanation. In some settings, faithfulness may be critical to ensure transparency. Lei et al. (2016) proposed a model to produce faithful rationales for neural text classification by defining independent snippet extraction and prediction modules. However, the discrete selection over input tokens performed by this method complicates training, leading to high variance and requiring careful hyperparameter tuning. We propose a simpler variant of this approach that provides faithful explanations by construction. In our scheme, named FRESH, arbitrary feature importance scores (e.g., gradients from a trained model) are used to induce binary labels over token inputs, which an extractor can be trained to predict. An independent classifier module is then trained exclusively on snippets provided by the extractor; these snippets thus constitute faithful explanations, even if the classifier is arbitrarily complex. In both automatic and manual evaluations we find that variants of this simple framework yield predictive performance superior to ‘end-to-end’ approaches, while being more general and easier to train. Code is available at https://github.com/successar/FRESH.

State-of-the-art models in NLP are now predominantly based on deep neural networks that are opaque in terms of how they come to make predictions. This limitation has increased interest in designing more interpretable deep models for NLP that reveal the ‘reasoning’ behind model outputs. But work in this direction has been conducted on different datasets and tasks with correspondingly unique aims and metrics; this makes it difficult to track progress. We propose the Evaluating Rationales And Simple English Reasoning (ERASER a benchmark to advance research on interpretable models in NLP. This benchmark comprises multiple datasets and tasks for which human annotations of “rationales” (supporting evidence) have been collected. We propose several metrics that aim to capture how well the rationales provided by models align with human rationales, and also how faithful these rationales are (i.e., the degree to which provided rationales influenced the corresponding predictions). Our hope is that releasing this benchmark facilitates progress on designing more interpretable NLP systems. The benchmark, code, and documentation are available at https://www.eraserbenchmark.com/

Natural language allows us to refer to novel composite concepts by combining expressions denoting their parts according to systematic rules, a property known as compositionality. In this paper, we study whether the language emerging in deep multi-agent simulations possesses a similar ability to refer to novel primitive combinations, and whether it accomplishes this feat by strategies akin to human-language compositionality. Equipped with new ways to measure compositionality in emergent languages inspired by disentanglement in representation learning, we establish three main results: First, given sufficiently large input spaces, the emergent language will naturally develop the ability to refer to novel composite concepts. Second, there is no correlation between the degree of compositionality of an emergent language and its ability to generalize. Third, while compositionality is not necessary for generalization, it provides an advantage in terms of language transmission: The more compositional a language is, the more easily it will be picked up by new learners, even when the latter differ in architecture from the original agents. We conclude that compositionality does not arise from simple generalization pressure, but if an emergent language does chance upon it, it will be more likely to survive and thrive.

In an era where generating content and publishing it is so easy, we are bombarded with information and are exposed to all kinds of claims, some of which do not always rank high on the truth scale. This paper suggests that the key to a longer-term, holistic, and systematic approach to navigating this information pollution is capturing the provenance of claims. To do that, we develop a formal definition of provenance graph for a given natural language claim, aiming to understand where the claim may come from and how it has evolved. To construct the graph, we model provenance inference, formulated mainly as an information extraction task and addressed via a textual entailment model. We evaluate our approach using two benchmark datasets, showing initial success in capturing the notion of provenance and its effectiveness on the application of claim verification.

Social biases are encoded in word embeddings. This presents a unique opportunity to study society historically and at scale, and a unique danger when embeddings are used in downstream applications. Here, we investigate the extent to which publicly-available word embeddings accurately reflect beliefs about certain kinds of people as measured via traditional survey methods. We find that biases found in word embeddings do, on average, closely mirror survey data across seventeen dimensions of social meaning. However, we also find that biases in embeddings are much more reflective of survey data for some dimensions of meaning (e.g. gender) than others (e.g. race), and that we can be highly confident that embedding-based measures reflect survey data only for the most salient biases.

A central concern in Computational Social Sciences (CSS) is fairness: where the role of NLP is to scale up text analysis to large corpora, the quality of automatic analyses should be as independent as possible of textual properties. We analyze the performance of a state-of-the-art neural model on the task of political claims detection (i.e., the identification of forward-looking statements made by political actors) and identify a strong frequency bias: claims made by frequent actors are recognized better. We propose two simple debiasing methods which mask proper names and pronouns during training of the model, thus removing personal information bias. We find that (a) these methods significantly decrease frequency bias while keeping the overall performance stable; and (b) the resulting models improve when evaluated in an out-of-domain setting.

Parody is a figurative device used to imitate an entity for comedic or critical purposes and represents a widespread phenomenon in social media through many popular parody accounts. In this paper, we present the first computational study of parody. We introduce a new publicly available data set of tweets from real politicians and their corresponding parody accounts. We run a battery of supervised machine learning models for automatically detecting parody tweets with an emphasis on robustness by testing on tweets from accounts unseen in training, across different genders and across countries. Our results show that political parody tweets can be predicted with an accuracy up to 90%. Finally, we identify the markers of parody through a linguistic analysis. Beyond research in linguistics and political communication, accurately and automatically detecting parody is important to improving fact checking for journalists and analytics such as sentiment analysis through filtering out parodical utterances.

The task of Dialogue Act Classification (DAC) that purports to capture communicative intent has been studied extensively. But these studies limit themselves to text. Non-verbal features (change of tone, facial expressions etc.) can provide cues to identify DAs, thus stressing the benefit of incorporating multi-modal inputs in the task. Also, the emotional state of the speaker has a substantial effect on the choice of the dialogue act, since conversations are often influenced by emotions. Hence, the effect of emotion too on automatic identification of DAs needs to be studied. In this work, we address the role of both multi-modality and emotion recognition (ER) in DAC. DAC and ER help each other by way of multi-task learning. One of the major contributions of this work is a new dataset- multimodal Emotion aware Dialogue Act dataset called EMOTyDA, collected from open-sourced dialogue datasets. To demonstrate the utility of EMOTyDA, we build an attention based (self, inter-modal, inter-task) multi-modal, multi-task Deep Neural Network (DNN) for joint learning of DAs and emotions. We show empirically that multi-modality and multi-tasking achieve better performance of DAC compared to uni-modal and single task DAC variants.

In this paper, we hypothesize that sarcasm is closely related to sentiment and emotion, and thereby propose a multi-task deep learning framework to solve all these three problems simultaneously in a multi-modal conversational scenario. We, at first, manually annotate the recently released multi-modal MUStARD sarcasm dataset with sentiment and emotion classes, both implicit and explicit. For multi-tasking, we propose two attention mechanisms, viz. Inter-segment Inter-modal Attention (Ie-Attention) and Intra-segment Inter-modal Attention (Ia-Attention). The main motivation of Ie-Attention is to learn the relationship between the different segments of the sentence across the modalities. In contrast, Ia-Attention focuses within the same segment of the sentence across the modalities. Finally, representations from both the attentions are concatenated and shared across the five classes (i.e., sarcasm, implicit sentiment, explicit sentiment, implicit emotion, explicit emotion) for multi-tasking. Experimental results on the extended version of the MUStARD dataset show the efficacy of our proposed approach for sarcasm detection over the existing state-of-the-art systems. The evaluation also shows that the proposed multi-task framework yields better performance for the primary task, i.e., sarcasm detection, with the help of two secondary tasks, emotion and sentiment analysis.

Multimodal Machine Translation (MMT) aims to introduce information from other modality, generally static images, to improve the translation quality. Previous works propose various incorporation methods, but most of them do not consider the relative importance of multiple modalities. Equally treating all modalities may encode too much useless information from less important modalities. In this paper, we introduce the multimodal self-attention in Transformer to solve the issues above in MMT. The proposed method learns the representation of images based on the text, which avoids encoding irrelevant information in images. Experiments and visualization analysis demonstrate that our model benefits from visual information and substantially outperforms previous works and competitive baselines in terms of various metrics.

In argumentation, people state premises to reason towards a conclusion. The conclusion conveys a stance towards some target, such as a concept or statement. Often, the conclusion remains implicit, though, since it is self-evident in a discussion or left out for rhetorical reasons. However, the conclusion is key to understanding an argument and, hence, to any application that processes argumentation. We thus study the question to what extent an argument’s conclusion can be reconstructed from its premises. In particular, we argue here that a decisive step is to infer a conclusion’s target, and we hypothesize that this target is related to the premises’ targets. We develop two complementary target inference approaches: one ranks premise targets and selects the top-ranked target as the conclusion target, the other finds a new conclusion target in a learned embedding space using a triplet neural network. Our evaluation on corpora from two domains indicates that a hybrid of both approaches is best, outperforming several strong baselines. According to human annotators, we infer a reasonably adequate conclusion target in 89% of the cases.