ACL 2022 Papers

Modern writing assistance applications are always equipped with a Grammatical Error Correction (GEC) model to correct errors in user-entered sentences. Different scenarios have varying requirements for correction behavior, e.g., performing more precise corrections (high precision) or providing more candidates for users (high recall). However, previous works adjust such trade-off only for sequence labeling approaches. In this paper, we propose a simple yet effective counterpart – Align-and-Predict Decoding (APD) for the most popular sequence-to-sequence models to offer more flexibility for the precision-recall trade-off. During inference, APD aligns the already generated sequence with input and adjusts scores of the following tokens. Experiments in both English and Chinese GEC benchmarks show that our approach not only adapts a single model to precision-oriented and recall-oriented inference, but also maximizes its potential to achieve state-of-the-art results. Our code is available at https://github.com/AutoTemp/Align-and-Predict.

We introduce a novel setup for low-resource task-oriented semantic parsing which incorporates several constraints that may arise in real-world scenarios: (1) lack of similar datasets/models from a related domain, (2) inability to sample useful logical forms directly from a grammar, and (3) privacy requirements for unlabeled natural utterances. Our goal is to improve a low-resource semantic parser using utterances collected through user interactions. In this highly challenging but realistic setting, we investigate data augmentation approaches involving generating a set of structured canonical utterances corresponding to logical forms, before simulating corresponding natural language and filtering the resulting pairs. We find that such approaches are effective despite our restrictive setup: in a low-resource setting on the complex SMCalFlow calendaring dataset (Andreas et al. 2020), we observe 33% relative improvement over a non-data-augmented baseline in top-1 match.

Current approaches to testing and debugging NLP models rely on highly variable human creativity and extensive labor, or only work for a very restrictive class of bugs. We present AdaTest, a process which uses large scale language models (LMs) in partnership with human feedback to automatically write unit tests highlighting bugs in a target model. Such bugs are then addressed through an iterative text-fix-retest loop, inspired by traditional software development. In experiments with expert and non-expert users and commercial / research models for 8 different tasks, AdaTest makes users 5-10x more effective at finding bugs than current approaches, and helps users effectively fix bugs without adding new bugs.

Neural coreference resolution models trained on one dataset may not transfer to new, low-resource domains. Active learning mitigates this problem by sampling a small subset of data for annotators to label. While active learning is well-defined for classification tasks, its application to coreference resolution is neither well-defined nor fully understood. This paper explores how to actively label coreference, examining sources of model uncertainty and document reading costs. We compare uncertainty sampling strategies and their advantages through thorough error analysis. In both synthetic and human experiments, labeling spans within the same document is more effective than annotating spans across documents. The findings contribute to a more realistic development of coreference resolution models.

Pre-trained language models have shown stellar performance in various downstream tasks. But, this usually comes at the cost of high latency and computation, hindering their usage in resource-limited settings. In this work, we propose a novel approach for reducing the computational cost of BERT with minimal loss in downstream performance. Our method dynamically eliminates less contributing tokens through layers, resulting in shorter lengths and consequently lower computational cost. To determine the importance of each token representation, we train a Contribution Predictor for each layer using a gradient-based saliency method. Our experiments on several diverse classification tasks show speedups up to 22x during inference time without much sacrifice in performance. We also validate the quality of the selected tokens in our method using human annotations in the ERASER benchmark. In comparison to other widely used strategies for selecting important tokens, such as saliency and attention, our proposed method has a significantly lower false positive rate in generating rationales. Our code is freely available at https://github.com/amodaresi/AdapLeR.

Recent machine reading comprehension datasets such as ReClor and LogiQA require performing logical reasoning over text. Conventional neural models are insufficient for logical reasoning, while symbolic reasoners cannot directly apply to text. To meet the challenge, we present a neural-symbolic approach which, to predict an answer, passes messages over a graph representing logical relations between text units. It incorporates an adaptive logic graph network (AdaLoGN) which adaptively infers logical relations to extend the graph and, essentially, realizes mutual and iterative reinforcement between neural and symbolic reasoning. We also implement a novel subgraph-to-node message passing mechanism to enhance context-option interaction for answering multiple-choice questions. Our approach shows promising results on ReClor and LogiQA.

Recent studies have shown the advantages of evaluating NLG systems using pairwise comparisons as opposed to direct assessment. Given k systems, a naive approach for identifying the top-ranked system would be to uniformly obtain pairwise comparisons from all k \choose 2 pairs of systems. However, this can be very expensive as the number of human annotations required would grow quadratically with k. In this work, we introduce Active Evaluation, a framework to efficiently identify the top-ranked system by actively choosing system pairs for comparison using dueling bandit algorithms. We perform extensive experiments with 13 dueling bandits algorithms on 13 NLG evaluation datasets spanning 5 tasks and show that the number of human annotations can be reduced by 80%. To further reduce the number of human annotations, we propose model-based dueling bandit algorithms which combine automatic evaluation metrics with human evaluations. Specifically, we eliminate sub-optimal systems even before the human annotation process and perform human evaluations only on test examples where the automatic metric is highly uncertain. This reduces the number of human annotations required further by 89%. In effect, we show that identifying the top-ranked system requires only a few hundred human annotations, which grow linearly with k. Lastly, we provide practical recommendations and best practices to identify the top-ranked system efficiently. Our code has been made publicly available at https://github.com/akashkm99/duelnlg

Evaluation of open-domain dialogue systems is highly challenging and development of better techniques is highlighted time and again as desperately needed. Despite substantial efforts to carry out reliable live evaluation of systems in recent competitions, annotations have been abandoned and reported as too unreliable to yield sensible results. This is a serious problem since automatic metrics are not known to provide a good indication of what may or may not be a high-quality conversation. Answering the distress call of competitions that have emphasized the urgent need for better evaluation techniques in dialogue, we present the successful development of human evaluation that is highly reliable while still remaining feasible and low cost. Self-replication experiments reveal almost perfectly repeatable results with a correlation of r=0.969. Furthermore, due to the lack of appropriate methods of statistical significance testing, the likelihood of potential improvements to systems occurring due to chance is rarely taken into account in dialogue evaluation, and the evaluation we propose facilitates application of standard tests. Since we have developed a highly reliable evaluation method, new insights into system performance can be revealed. We therefore include a comparison of state-of-the-art models (i) with and without personas, to measure the contribution of personas to conversation quality, as well as (ii) prescribed versus freely chosen topics. Interestingly with respect to personas, results indicate that personas do not positively contribute to conversation quality as expected.

A limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses, primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge. One way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response. In this paper, we propose a post-hoc knowledge-injection technique where we first retrieve a diverse set of relevant knowledge snippets conditioned on both the dialog history and an initial response from an existing dialog model. We construct multiple candidate responses, individually injecting each retrieved snippet into the initial response using a gradient-based decoding method, and then select the final response with an unsupervised ranking step. Our experiments in goal-oriented and knowledge-grounded dialog settings demonstrate that human annotators judge the outputs from the proposed method to be more engaging and informative compared to responses from prior dialog systems. We further show that knowledge-augmentation promotes success in achieving conversational goals in both experimental settings.

Online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded. Good online alignments facilitate important applications such as lexically constrained translation where user-defined dictionaries are used to inject lexical constraints into the translation model. We propose a novel posterior alignment technique that is truly online in its execution and superior in terms of alignment error rates compared to existing methods. Our proposed inference technique jointly considers alignment and token probabilities in a principled manner and can be seamlessly integrated within existing constrained beam-search decoding algorithms. On five language pairs, including two distant language pairs, we achieve consistent drop in alignment error rates. When deployed on seven lexically constrained translation tasks, we achieve significant improvements in BLEU specifically around the constrained positions.

Code search is to search reusable code snippets from source code corpus based on natural languages queries. Deep learning-based methods on code search have shown promising results. However, previous methods focus on retrieval accuracy, but lacked attention to the efficiency of the retrieval process. We propose a novel method CoSHC to accelerate code search with deep hashing and code classification, aiming to perform efficient code search without sacrificing too much accuracy. To evaluate the effectiveness of CoSHC, we apply our methodon five code search models. Extensive experimental results indicate that compared with previous code search baselines, CoSHC can save more than 90% of retrieval time meanwhile preserving at least 99% of retrieval accuracy.

Transformers have recently been shown to be capable of reliably performing logical reasoning over facts and rules expressed in natural language, but abductive reasoning - inference to the best explanation of an unexpected observation - has been underexplored despite significant applications to scientific discovery, common-sense reasoning, and model interpretability. This paper presents AbductionRules, a group of natural language datasets designed to train and test generalisable abduction over natural-language knowledge bases. We use these datasets to finetune pretrained Transformers and discuss their performance, finding that our models learned generalisable abductive techniques but also learned to exploit the structure of our data. Finally, we discuss the viability of this approach to abductive reasoning and ways in which it may be improved in future work.

The MultiWOZ 2.0 dataset has greatly boosted the research on dialogue state tracking (DST). However, substantial noise has been discovered in its state annotations. Such noise brings about huge challenges for training DST models robustly. Although several refined versions, including MultiWOZ 2.1-2.4, have been published recently, there are still lots of noisy labels, especially in the training set. Besides, it is costly to rectify all the problematic annotations. In this paper, instead of improving the annotation quality further, we propose a general framework, named ASSIST (lAbel noiSe-robuSt dIalogue State Tracking), to train DST models robustly from noisy labels. ASSIST first generates pseudo labels for each sample in the training set by using an auxiliary model trained on a small clean dataset, then puts the generated pseudo labels and vanilla noisy labels together to train the primary model. We show the validity of ASSIST theoretically. Experimental results also demonstrate that ASSIST improves the joint goal accuracy of DST by up to 28.16% on MultiWOZ 2.0 and 8.41% on MultiWOZ 2.4, compared to using only the vanilla noisy labels.

Generic summaries try to cover an entire document and query-based summaries try to answer document-specific questions. But real users’ needs often fall in between these extremes and correspond to aspects, high-level topics discussed among similar types of documents. In this paper, we collect a dataset of realistic aspect-oriented summaries, AspectNews, which covers different subtopics about articles in news sub-domains. We annotate data across two domains of articles, earthquakes and fraud investigations, where each article is annotated with two distinct summaries focusing on different aspects for each domain. A system producing a single generic summary cannot concisely satisfy both aspects. Our focus in evaluation is how well existing techniques can generalize to these domains without seeing in-domain training data, so we turn to techniques to construct synthetic training data that have been used in query-focused summarization work. We compare several training schemes that differ in how strongly keywords are used and how oracle summaries are extracted. Our evaluation shows that our final approach yields (a) focused summaries, better than those from a generic summarization system or from keyword matching; (b) a system sensitive to the choice of keywords.

Prompt-based learning, which exploits knowledge from pre-trained language models by providing textual prompts and designing appropriate answer-category mapping methods, has achieved impressive successes on few-shot text classification and natural language inference (NLI). Because of the diverse linguistic expression, there exist many answer tokens for the same category. However, both manual answer design and automatic answer search constrain answer space and therefore hardly achieve ideal performance. To address this issue, we propose an answer space clustered prompting model (ASCM) together with a synonym initialization method (SI) which automatically categorizes all answer tokens in a semantic-clustered embedding space. We also propose a stable semi-supervised method named stair learning (SL) that orderly distills knowledge from better models to weaker models. Extensive experiments demonstrate that our ASCM+SL significantly outperforms existing state-of-the-art techniques in few-shot settings.

Abstract Meaning Representation (AMR) is a semantic representation for NLP/NLU. In this paper, we propose to use it for data augmentation in NLP. Our proposed data augmentation technique, called AMR-DA, converts a sample sentence to an AMR graph, modifies the graph according to various data augmentation policies, and then generates augmentations from graphs. Our method combines both sentence-level techniques like back translation and token-level techniques like EDA (Easy Data Augmentation). To evaluate the effectiveness of our method, we apply it to the tasks of semantic textual similarity (STS) and text classification. For STS, our experiments show that AMR-DA boosts the performance of the state-of-the-art models on several STS benchmarks. For text classification, AMR-DA outperforms EDA and AEDA and leads to more robust improvements.

Transformer architectures have achieved state- of-the-art results on a variety of natural language processing (NLP) tasks. However, their attention mechanism comes with a quadratic complexity in sequence lengths, making the computational overhead prohibitive, especially for long sequences. Attention context can be seen as a random-access memory with each token taking a slot. Under this perspective, the memory size grows linearly with the sequence length, and so does the overhead of reading from it. One way to improve the efficiency is to bound the memory size. We show that disparate approaches can be subsumed into one abstraction, attention with bounded-memory control (ABC), and they vary in their organization of the memory. ABC reveals new, unexplored possibilities. First, it connects several efficient attention variants that would otherwise seem apart. Second, this abstraction gives new insights—an established approach (Wang et al., 2020b) previously thought to not be applicable in causal attention, actually is. Last, we present a new instance of ABC, which draws inspiration from existing ABC approaches, but replaces their heuristic memory-organizing functions with a learned, contextualized one. Our experiments on language modeling, machine translation, and masked language model finetuning show that our approach outperforms previous efficient attention models; compared to the strong transformer baselines, it significantly improves the inference time and space efficiency with no or negligible accuracy loss.

We propose Composition Sampling, a simple but effective method to generate diverse outputs for conditional generation of higher quality compared to previous stochastic decoding strategies. It builds on recently proposed plan-based neural generation models (FROST, Narayan et al, 2021) that are trained to first create a composition of the output and then generate by conditioning on it and the input. Our approach avoids text degeneration by first sampling a composition in the form of an entity chain and then using beam search to generate the best possible text grounded to this entity chain. Experiments on summarization (CNN/DailyMail and XSum) and question generation (SQuAD), using existing and newly proposed automaticmetrics together with human-based evaluation, demonstrate that Composition Sampling is currently the best available decoding strategy for generating diverse meaningful outputs.

The goal of the cross-lingual summarization (CLS) is to convert a document in one language (e.g., English) to a summary in another one (e.g., Chinese). The CLS task is essentially the combination of machine translation (MT) and monolingual summarization (MS), and thus there exists the hierarchical relationship between MT&MS and CLS. Existing studies on CLS mainly focus on utilizing pipeline methods or jointly training an end-to-end model through an auxiliary MT or MS objective. However, it is very challenging for the model to directly conduct CLS as it requires both the abilities to translate and summarize. To address this issue, we propose a hierarchical model for the CLS task, based on the conditional variational auto-encoder. The hierarchical model contains two kinds of latent variables at the local and global levels, respectively. At the local level, there are two latent variables, one for translation and the other for summarization. As for the global level, there is another latent variable for cross-lingual summarization conditioned on the two local-level variables. Experiments on two language directions (English-Chinese) verify the effectiveness and superiority of the proposed approach. In addition, we show that our model is able to generate better cross-lingual summaries than comparison models in the few-shot setting.