| Total: 139
A snowclone is a customizable phrasal template that can be realized in multiple, instantly recognized variants. For example, “* is the new *" (Orange is the new black, 40 is the new 30). Snowclones are extensively used in social media. In this paper, we study snowclones originating from pop-culture quotes; our goal is to automatically detect cultural references in text. We introduce a new, publicly available data set of pop-culture quotes and their corresponding snowclone usages and train models on them. We publish code for Catchphrase, an internet browser plugin to automatically detect and mark references in real-time, and examine its performance via a user study. Aside from assisting people to better comprehend cultural references, we hope that detecting snowclones can complement work on paraphrasing and help tackling long-standing questions in social science about the dynamics of information propagation.
Large-scale pretrained language models have led to dramatic improvements in text generation. Impressive performance can be achieved by finetuning only on a small number of instances (few-shot setting). Nonetheless, almost all previous work simply applies random sampling to select the few-shot training instances. Little to no attention has been paid to the selection strategies and how they would affect model performance. In this work, we present a study on training instance selection in few-shot neural text generation. The selection decision is made based only on the unlabeled data so as to identify the most worthwhile data points that should be annotated under some budget of labeling cost. Based on the intuition that the few-shot training instances should be diverse and representative of the entire data distribution, we propose a simple selection strategy with K-means clustering. We show that even with the naive clustering-based approach, the generation models consistently outperform random sampling on three text generation tasks: data-to-text generation, document summarization and question generation. The code and training data are made available. We hope that this work will call for more attention on this largely unexplored area.
The introduction of pretrained language models has reduced many complex task-specific NLP models to simple lightweight layers. An exception to this trend is coreference resolution, where a sophisticated task-specific model is appended to a pretrained transformer encoder. While highly effective, the model has a very large memory footprint – primarily due to dynamically-constructed span and span-pair representations – which hinders the processing of complete documents and the ability to train on multiple instances in a single batch. We introduce a lightweight end-to-end coreference model that removes the dependency on span representations, handcrafted features, and heuristics. Our model performs competitively with the current standard model, while being simpler and more efficient.
In comparison with English, due to the lack of explicit word boundary and tenses information, Chinese Named Entity Recognition (NER) is much more challenging. In this paper, we propose a boundary enhanced approach for better Chinese NER. In particular, our approach enhances the boundary information from two perspectives. On one hand, we enhance the representation of the internal dependency of phrases by an additional Graph Attention Network(GAT) layer. On the other hand, taking the entity head-tail prediction (i.e., boundaries) as an auxiliary task, we propose an unified framework to learn the boundary information and recognize the NE jointly. Experiments on both the OntoNotes and the Weibo corpora show the effectiveness of our approach.
The high-quality translation results produced by machine translation (MT) systems still pose a huge challenge for automatic evaluation. Current MT evaluation pays the same attention to each sentence component, while the questions of real-world examinations (e.g., university examinations) have different difficulties and weightings. In this paper, we propose a novel difficulty-aware MT evaluation metric, expanding the evaluation dimension by taking translation difficulty into consideration. A translation that fails to be predicted by most MT systems will be treated as a difficult one and assigned a large weight in the final score function, and conversely. Experimental results on the WMT19 English-German Metrics shared tasks show that our proposed method outperforms commonly used MT metrics in terms of human correlation. In particular, our proposed method performs well even when all the MT systems are very competitive, which is when most existing metrics fail to distinguish between them. The source code is freely available at https://github.com/NLP2CT/Difficulty-Aware-MT-Evaluation.
Humor recognition has been widely studied as a text classification problem using data-driven approaches. However, most existing work does not examine the actual joke mechanism to understand humor. We break down any joke into two distinct components: the set-up and the punchline, and further explore the special relationship between them. Inspired by the incongruity theory of humor, we model the set-up as the part developing semantic uncertainty, and the punchline disrupting audience expectations. With increasingly powerful language models, we were able to feed the set-up along with the punchline into the GPT-2 language model, and calculate the uncertainty and surprisal values of the jokes. By conducting experiments on the SemEval 2021 Task 7 dataset, we found that these two features have better capabilities of telling jokes from non-jokes, compared with existing baselines.
Disentanglement of latent representations into content and style spaces has been a commonly employed method for unsupervised text style transfer. These techniques aim to learn the disentangled representations and tweak them to modify the style of a sentence. In this paper, we propose a counterfactual-based method to modify the latent representation, by posing a ‘what-if’ scenario. This simple and disciplined approach also enables a fine-grained control on the transfer strength. We conduct experiments with the proposed methodology on multiple attribute transfer tasks like Sentiment, Formality and Excitement to support our hypothesis.
Shapley Values, a solution to the credit assignment problem in cooperative game theory, are a popular type of explanation in machine learning, having been used to explain the importance of features, embeddings, and even neurons. In NLP, however, leave-one-out and attention-based explanations still predominate. Can we draw a connection between these different methods? We formally prove that — save for the degenerate case — attention weights and leave-one-out values cannot be Shapley Values. Attention flow is a post-processed variant of attention weights obtained by running the max-flow algorithm on the attention graph. Perhaps surprisingly, we prove that attention flows are indeed Shapley Values, at least at the layerwise level. Given the many desirable theoretical qualities of Shapley Values — which has driven their adoption among the ML community — we argue that NLP practitioners should, when possible, adopt attention flow explanations alongside more traditional ones.
Video paragraph captioning aims to generate a set of coherent sentences to describe a video that contains several events. Most previous methods simplify this task by using ground-truth event segments. In this work, we propose a novel framework by taking this task as a text summarization task. We first generate lots of sentence-level captions focusing on different video clips and then summarize these captions to obtain the final paragraph caption. Our method does not depend on ground-truth event segments. Experiments on two popular datasets ActivityNet Captions and YouCookII demonstrate the advantages of our new framework. On the ActivityNet dataset, our method even outperforms some previous methods using ground-truth event segment labels.
Deep learning algorithms have shown promising results in visual question answering (VQA) tasks, but a more careful look reveals that they often do not understand the rich signal they are being fed with. To understand and better measure the generalization capabilities of VQA systems, we look at their robustness to counterfactually augmented data. Our proposed augmentations are designed to make a focused intervention on a specific property of the question such that the answer changes. Using these augmentations, we propose a new robustness measure, Robustness to Augmented Data (RAD), which measures the consistency of model predictions between original and augmented examples. Through extensive experimentation, we show that RAD, unlike classical accuracy measures, can quantify when state-of-the-art systems are not robust to counterfactuals. We find substantial failure cases which reveal that current VQA systems are still brittle. Finally, we connect between robustness and generalization, demonstrating the predictive power of RAD for performance on unseen augmentations.
Existing approaches for the Table-to-Text task suffer from issues such as missing information, hallucination and repetition. Many approaches to this problem use Reinforcement Learning (RL), which maximizes a single manually defined reward, such as BLEU. In this work, we instead pose the Table-to-Text task as Inverse Reinforcement Learning (IRL) problem. We explore using multiple interpretable unsupervised reward components that are combined linearly to form a composite reward function. The composite reward function and the description generator are learned jointly. We find that IRL outperforms strong RL baselines marginally. We further study the generalization of learned IRL rewards in scenarios involving domain adaptation. Our experiments reveal significant challenges in using IRL for this task.
Most fact checking models for automatic fake news detection are based on reasoning: given a claim with associated evidence, the models aim to estimate the claim veracity based on the supporting or refuting content within the evidence. When these models perform well, it is generally assumed to be due to the models having learned to reason over the evidence with regards to the claim. In this paper, we investigate this assumption of reasoning, by exploring the relationship and importance of both claim and evidence. Surprisingly, we find on political fact checking datasets that most often the highest effectiveness is obtained by utilizing only the evidence, as the impact of including the claim is either negligible or harmful to the effectiveness. This highlights an important problem in what constitutes evidence in existing approaches for automatic fake news detection.
Despite end-to-end neural systems making significant progress in the last decade for task-oriented as well as chit-chat based dialogue systems, most dialogue systems rely on hybrid approaches which use a combination of rule-based, retrieval and generative approaches for generating a set of ranked responses. Such dialogue systems need to rely on a fallback mechanism to respond to out-of-domain or novel user queries which are not answerable within the scope of the dialogue system. While, dialogue systems today rely on static and unnatural responses like “I don’t know the answer to that question” or “I’m not sure about that”, we design a neural approach which generates responses which are contextually aware with the user query as well as say no to the user. Such customized responses provide paraphrasing ability and contextualization as well as improve the interaction with the user and reduce dialogue monotonicity. Our simple approach makes use of rules over dependency parses and a text-to-text transformer fine-tuned on synthetic data of question-response pairs generating highly relevant, grammatical as well as diverse questions. We perform automatic and manual evaluations to demonstrate the efficacy of the system.
Spoken Language Understanding (SLU) systems parse speech into semantic structures like dialog acts and slots. This involves the use of an Automatic Speech Recognizer (ASR) to transcribe speech into multiple text alternatives (hypotheses). Transcription errors, ordinary in ASRs, impact downstream SLU performance negatively. Common approaches to mitigate such errors involve using richer information from the ASR, either in form of N-best hypotheses or word-lattices. We hypothesize that transformer models will learn better with a simpler utterance representation using the concatenation of the N-best ASR alternatives, where each alternative is separated by a special delimiter [SEP]. In our work, we test our hypothesis by using the concatenated N-best ASR alternatives as the input to the transformer encoder models, namely BERT and XLM-RoBERTa, and achieve equivalent performance to the prior state-of-the-art model on DSTC2 dataset. We also show that our approach significantly outperforms the prior state-of-the-art when subjected to the low data regime. Additionally, this methodology is accessible to users of third-party ASR APIs which do not provide word-lattice information.
Is bias amplified when neural machine translation (NMT) models are optimized for speed and evaluated on generic test sets using BLEU? We investigate architectures and techniques commonly used to speed up decoding in Transformer-based models, such as greedy search, quantization, average attention networks (AANs) and shallow decoder models and show their effect on gendered noun translation. We construct a new gender bias test set, SimpleGEN, based on gendered noun phrases in which there is a single, unambiguous, correct answer. While we find minimal overall BLEU degradation as we apply speed optimizations, we observe that gendered noun translation performance degrades at a much faster rate.
State-of-the-art machine translation (MT) systems are typically trained to generate “standard” target language; however, many languages have multiple varieties (regional varieties, dialects, sociolects, non-native varieties) that are different from the standard language. Such varieties are often low-resource, and hence do not benefit from contemporary NLP solutions, MT included. We propose a general framework to rapidly adapt MT systems to generate language varieties that are close to, but different from, the standard target language, using no parallel (source–variety) data. This also includes adaptation of MT systems to low-resource typologically-related target languages. We experiment with adapting an English–Russian MT system to generate Ukrainian and Belarusian, an English–Norwegian Bokmål system to generate Nynorsk, and an English–Arabic system to generate four Arabic dialects, obtaining significant improvements over competitive baselines.
Sparse attention has been claimed to increase model interpretability under the assumption that it highlights influential inputs. Yet the attention distribution is typically over representations internal to the model rather than the inputs themselves, suggesting this assumption may not have merit. We build on the recent work exploring the interpretability of attention; we design a set of experiments to help us understand how sparsity affects our ability to use attention as an explainability tool. On three text classification tasks, we verify that only a weak relationship between inputs and co-indexed intermediate representations exists—under sparse attention and otherwise. Further, we do not find any plausible mappings from sparse attention distributions to a sparse set of influential inputs through other avenues. Rather, we observe in this setting that inducing sparsity may make it less plausible that attention can be used as a tool for understanding model behavior.
Mechanisms for encoding positional information are central for transformer-based language models. In this paper, we analyze the position embeddings of existing language models, finding strong evidence of translation invariance, both for the embeddings themselves and for their effect on self-attention. The degree of translation invariance increases during training and correlates positively with model performance. Our findings lead us to propose translation-invariant self-attention (TISA), which accounts for the relative position between tokens in an interpretable fashion without needing conventional position embeddings. Our proposal has several theoretical advantages over existing position-representation approaches. Proof-of-concept experiments show that it improves on regular ALBERT on GLUE tasks, while only adding orders of magnitude less positional parameters.
Determining the relative importance of the elements in a sentence is a key factor for effortless natural language understanding. For human language processing, we can approximate patterns of relative importance by measuring reading fixations using eye-tracking technology. In neural language models, gradient-based saliency methods indicate the relative importance of a token for the target objective. In this work, we compare patterns of relative importance in English language processing by humans and models and analyze the underlying linguistic patterns. We find that human processing patterns in English correlate strongly with saliency-based importance in language models and not with attention-based importance. Our results indicate that saliency could be a cognitively more plausible metric for interpreting neural language models. The code is available on github: https://github.com/beinborn/relative_importance.
Pretrained language models (PLM) achieve surprising performance on the Choice of Plausible Alternatives (COPA) task. However, whether PLMs have truly acquired the ability of causal reasoning remains a question. In this paper, we investigate the problem of semantic similarity bias and reveal the vulnerability of current COPA models by certain attacks. Previous solutions that tackle the superficial cues of unbalanced token distribution still encounter the same problem of semantic bias, even more seriously due to the utilization of more training data. We mitigate this problem by simply adding a regularization loss and experimental results show that this solution not only improves the model’s generalization ability, but also assists the models to perform more robustly on a challenging dataset, BCOPA-CE, which has unbiased token distribution and is more difficult for models to distinguish cause and effect.
A current open question in natural language processing is to what extent language models, which are trained with access only to the form of language, are able to capture the meaning of language. This question is challenging to answer in general, as there is no clear line between meaning and form, but rather meaning constrains form in consistent ways. The goal of this study is to offer insights into a narrower but critical subquestion: Under what conditions should we expect that meaning and form covary sufficiently, such that a language model with access only to form might nonetheless succeed in emulating meaning? Focusing on several formal languages (propositional logic and a set of programming languages), we generate training corpora using a variety of motivated constraints, and measure a distributional language model’s ability to differentiate logical symbols (AND, OR, and NOT). Our findings are largely negative: none of our simulated training corpora result in models which definitively differentiate meaningfully different symbols (e.g., AND vs. OR), suggesting a limitation to the types of semantic signals that current models are able to exploit.
The predominant challenge in weakly supervised semantic parsing is that of spurious programs that evaluate to correct answers for the wrong reasons. Prior work uses elaborate search strategies to mitigate the prevalence of spurious programs; however, they typically consider only one input at a time. In this work we explore the use of consistency between the output programs for related inputs to reduce the impact of spurious programs. We bias the program search (and thus the model’s training signal) towards programs that map the same phrase in related inputs to the same sub-parts in their respective programs. Additionally, we study the importance of designing logical formalisms that facilitate this kind of consistency-based training. We find that a more consistent formalism leads to improved model performance even without consistency-based training. When combined together, these two insights lead to a 10% absolute improvement over the best prior result on the Natural Language Visual Reasoning dataset.
In this paper, we present an improved model for voicing silent speech, where audio is synthesized from facial electromyography (EMG) signals. To give our model greater flexibility to learn its own input features, we directly use EMG signals as input in the place of hand-designed features used by prior work. Our model uses convolutional layers to extract features from the signals and Transformer layers to propagate information across longer distances. To provide better signal for learning, we also introduce an auxiliary task of predicting phoneme labels in addition to predicting speech audio features. On an open vocabulary intelligibility evaluation, our model improves the state of the art for this task by an absolute 25.8%.
Whereas much of the success of the current generation of neural language models has been driven by increasingly large training corpora, relatively little research has been dedicated to analyzing these massive sources of textual data. In this exploratory analysis, we delve deeper into the Common Crawl, a colossal web corpus that is extensively used for training language models. We find that it contains a significant amount of undesirable content, including hate speech and sexually explicit content, even after filtering procedures. We discuss the potential impacts of this content on language models and conclude with future research directions and a more mindful approach to corpus collection and analysis.
Most current quality estimation (QE) models for machine translation are trained and evaluated in a static setting where training and test data are assumed to be from a fixed distribution. However, in real-life settings, the test data that a deployed QE model would be exposed to may differ from its training data. In particular, training samples are often labelled by one or a small set of annotators, whose perceptions of translation quality and needs may differ substantially from those of end-users, who will employ predictions in practice. To address this challenge, we propose an online Bayesian meta-learning framework for the continuous training of QE models that is able to adapt them to the needs of different users, while being robust to distributional shifts in training and test data. Experiments on data with varying number of users and language characteristics validate the effectiveness of the proposed approach.