Despite the recent success of multi-task learning and pre-finetuning for natural language understanding, few works have studied the effects of task families on abstractive text summarization. Task families are a form of task grouping during the pre-finetuning stage to learn common skills, such as reading comprehension. To close this gap, we analyze the influence of multi-task learning strategies using task families for the English abstractive text summarization task. We group tasks into one of three strategies, i.e., sequential, simultaneous, and continual multi-task learning, and evaluate trained models through two downstream tasks. We find that certain combinations of task families (e.g., advanced reading comprehension and natural language inference) positively impact downstream performance. Further, we find that choice and combinations of task families influence downstream performance more than the training scheme, supporting the use of task families for abstractive text summarization.
Social media has increasingly played a key role in emergency response: first responders can use public posts to better react to ongoing crisis events and deploy the necessary resources where they are most needed. Timeline extraction and abstractive summarization are critical technical tasks to leverage large numbers of social media posts about events. Unfortunately, there are few datasets for benchmarking technical approaches for those tasks. This paper presents CrisisLTLSum, the largest dataset of local crisis event timelines available to date. CrisisLTLSum contains 1,000 crisis event timelines across four domains: wildfires, local fires, traffic, and storms. We built CrisisLTLSum using a semi-automated cluster-then-refine approach to collect data from the public Twitter stream. Our initial experiments indicate a significant gap between the performance of strong baselines compared to the human performance on both tasks. Our dataset, code, and models are publicly available.
Abstractive summarization models often generate inconsistent summaries containing factual errors or hallucinated content. Recent works focus on correcting factual errors in generated summaries via post-editing. Such correction models are trained using adversarial non-factual summaries constructed using heuristic rules for injecting errors. However, generating non-factual summaries using heuristics often does not generalize well to actual model errors. In this work, we propose to generate hard, representative synthetic examples of non-factual summaries through infilling language models. With this data, we train a more robust fact-correction model to post-edit the summaries to improve factual consistency. Through quantitative and qualitative experiments on two popular summarization datasets -- CNN/DM and XSum -- we show that our approach vastly outperforms prior methods in correcting erroneous summaries. Our model -- FactEdit -- improves factuality scores by over ~11 points on CNN/DM and over ~31 points on XSum on average across multiple summarization models, producing more factual summaries while maintaining competitive summarization quality.
Despite significant progress in the quality of language generated from abstractive summarization models, these models still exhibit the tendency to hallucinate, i.e., output content not supported by the source document. A number of works have tried to fix--or at least uncover the source of--the problem with limited success. In this paper, we identify a simple criterion under which models are significantly more likely to assign more probability to hallucinated content during generation: high model uncertainty. This finding offers a potential explanation for hallucinations: models default to favoring text with high marginal probability, i.e., high-frequency occurrences in the training set, when uncertain about a continuation. It also motivates possible routes for real-time intervention during decoding to prevent such hallucinations. We propose a decoding strategy that switches to optimizing for pointwise mutual information of the source and target token--rather than purely the probability of the target token--when the model exhibits uncertainty. Experiments on the XSum dataset show that our method decreases the probability of hallucinated tokens while maintaining the Rouge and BertS scores of top-performing decoding strategies.
Abstractive summarization models typically learn to capture the salient information from scratch implicitly. Recent literature adds extractive summaries as guidance for abstractive summarization models to provide hints of salient content and achieves better performance. However, extractive summaries as guidance could be over strict, leading to information loss or noisy signals. Furthermore, it cannot easily adapt to documents with various abstractiveness. As the number and allocation of salience content pieces vary, it is hard to find a fixed threshold deciding which content should be included in the guidance. In this paper, we propose a novel summarization approach with a flexible and reliable salience guidance, namely SEASON (SaliencE Allocation as Guidance for Abstractive SummarizatiON). SEASON utilizes the allocation of salience expectation to guide abstractive summarization and adapts well to articles in different abstractiveness. Automatic and human evaluations on two benchmark datasets show that the proposed method is effective and reliable. Empirical results on more than one million news articles demonstrate a natural fifteen-fifty salience split for news article sentences, providing a useful insight for composing news articles.
Sequence-to-sequence deep neural models fine-tuned for abstractive summarization can achieve great performance on datasets with enough human annotations. Yet, it has been shown that they have not reached their full potential, with a wide gap between the top beam search output and the oracle beam. Recently, re-ranking methods have been proposed, to learn to select a better summary candidate. However, such methods are limited by the summary quality aspects captured by the first-stage candidates. To bypass this limitation, we propose a new paradigm in second-stage abstractive summarization called SummaFusion that fuses several summary candidates to produce a novel abstractive second-stage summary. Our method works well on several summarization datasets, improving both the ROUGE scores and qualitative properties of fused summaries. It is especially good when the candidates to fuse are worse, such as in the few-shot setup where we set a new state-of-the-art. We will make our code and checkpoints available at https://github.com/ntunlp/SummaFusion/.
Summarization of legal case judgement documents is a challenging problem in Legal NLP. However, not much analyses exist on how different families of summarization models (e.g., extractive vs. abstractive) perform when applied to legal case documents. This question is particularly important since many recent transformer-based abstractive summarization models have restrictions on the number of input tokens, and legal documents are known to be very long. Also, it is an open question on how best to evaluate legal case document summarization systems. In this paper, we carry out extensive experiments with several extractive and abstractive summarization methods (both supervised and unsupervised) over three legal summarization datasets that we have developed. Our analyses, that includes evaluation by law practitioners, lead to several interesting insights on legal summarization in specific and long document summarization in general.
Saliency maps can explain a neural model's predictions by identifying important input features. They are difficult to interpret for laypeople, especially for instances with many features. In order to make them more accessible, we formalize the underexplored task of translating saliency maps into natural language and compare methods that address two key challenges of this approach -- what and how to verbalize. In both automatic and human evaluation setups, using token-level attributions from text classification tasks, we compare two novel methods (search-based and instruction-based verbalizations) against conventional feature importance representations (heatmap visualizations and extractive rationales), measuring simulatability, faithfulness, helpfulness and ease of understanding. Instructing GPT-3.5 to generate saliency map verbalizations yields plausible explanations which include associations, abstractive summarization and commonsense reasoning, achieving by far the highest human ratings, but they are not faithfully capturing numeric information and are inconsistent in their interpretation of the task. In comparison, our search-based, model-free verbalization approach efficiently completes templated verbalizations, is faithful by design, but falls short in helpfulness and simulatability. Our results suggest that saliency map verbalization makes feature attribution explanations more comprehensible and less cognitively challenging to humans than conventional representations.
Despite recent progress in abstractive summarization, models often generate summaries with factual errors. Numerous approaches to detect these errors have been proposed, the most popular of which are question answering (QA)-based factuality metrics. These have been shown to work well at predicting summary-level factuality and have potential to localize errors within summaries, but this latter capability has not been systematically evaluated in past research. In this paper, we conduct the first such analysis and find that, contrary to our expectations, QA-based frameworks fail to correctly identify error spans in generated summaries and are outperformed by trivial exact match baselines. Our analysis reveals a major reason for such poor localization: questions generated by the QG module often inherit errors from non-factual summaries which are then propagated further into downstream modules. Moreover, even human-in-the-loop question generation cannot easily offset these problems. Our experiments conclusively show that there exist fundamental issues with localization using the QA framework which cannot be fixed solely by stronger QA and QG models.
The issue of factual consistency in abstractive summarization has received extensive attention in recent years, and the evaluation of factual consistency between summary and document has become an important and urgent task. Most of the current evaluation metrics are adopted from the question answering (QA) or natural language inference (NLI) task. However, the application of QA-based metrics is extremely time-consuming in practice while NLI-based metrics are lack of interpretability. In this paper, we propose a cloze-based evaluation framework called ClozE and show the great potential of the cloze-based metric. It inherits strong interpretability from QA, while maintaining the speed of NLI- level reasoning. We demonstrate that ClozE can reduce the evaluation time by nearly 96% relative to QA-based metrics while retaining their interpretability and performance through experiments on six human-annotated datasets and a meta-evaluation benchmark GO FIGURE (Gabriel et al., 2021). Finally, we discuss three important facets of ClozE in practice, which further shows better overall performance of ClozE compared to other metrics.
Despite the success achieved in neural abstractive summarization based on pre-trained language models, one unresolved issue is that the generated summaries are not always faithful to the input document. There are two possible causes of the unfaithfulness problem: (1) the summarization model fails to understand or capture the gist of the input text, and (2) the model over-relies on the language model to generate fluent but inadequate words. In this work, we propose a Faithfulness Enhanced Summarization model (FES), which is designed for addressing these two problems and improving faithfulness in abstractive summarization. For the first problem, we propose to use question-answering (QA) to examine whether the encoder fully grasps the input document and can answer the questions on the key information in the input. The QA attention on the proper input words can also be used to stipulate how the decoder should attend to the source. For the second problem, we introduce a max-margin loss defined on the difference between the language and the summarization model, aiming to prevent the overconfidence of the language model. Extensive experiments on two benchmark summarization datasets, CNN/DM and XSum, demonstrate that our model significantly outperforms strong baselines. The evaluation of factual consistency also shows that our model generates more faithful summaries than baselines.
Abstractive text summarization has recently become a popular approach, but data hallucination remains a serious problem, including with quantitative data. We propose a set of probing tests to evaluate the efficacy of abstract summarization models' modeling of quantitative values found in the input text. Our results show that in most cases, the encoders of recent SOTA-performing models struggle to provide embeddings that adequately represent quantitative values in the input compared to baselines, and in particular, they outperform random representations in some, but surprisingly not all, cases. Under our assumptions, this suggests that the encoder's performance contributes to the quantity hallucination problem. One model type in particular, DistilBART-CDM, was observed to underperform randomly initialized representations for several experiments, and performance versus BERT suggests that standard pretraining and fine-tuning approaches for the summarization task may play a role in underperformance for some encoders.
Conditional language models are predominantly trained with maximum likelihood estimation (MLE), giving probability mass to sparsely observed target sequences. While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. This has been empirically observed in beam search decoding as output quality degrading with large beam sizes, and decoding strategies benefiting from heuristics such as length normalization and repetition-blocking. In this work, we introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model's latent space. With SLiC, decoding heuristics become unnecessary and decoding candidates' quality significantly improves regardless of the decoding method. Furthermore, SLiC shows no sign of diminishing returns with model scale, and presents alternative ways to improve quality with limited training and inference budgets. With SLiC, we exceed or match SOTA results on a wide range of generation tasks spanning abstractive summarization, question generation, abstractive question answering and data-to-text generation, even with modest-sized models.
Machine learning algorithms typically assume independent and identically distributed samples in training and at test time. Much work has shown that high-performing ML classifiers can degrade significantly and provide overly-confident, wrong classification predictions, particularly for out-of-distribution (OOD) inputs. Conditional language models (CLMs) are predominantly trained to classify the next token in an output sequence, and may suffer even worse degradation on OOD inputs as the prediction is done auto-regressively over many steps. Furthermore, the space of potential low-quality outputs is larger as arbitrary text can be generated and it is important to know when to trust the generated output. We present a highly accurate and lightweight OOD detection method for CLMs, and demonstrate its effectiveness on abstractive summarization and translation. We also show how our method can be used under the common and realistic setting of distribution shift for selective generation (analogous to selective prediction for classification) of high-quality outputs, while automatically abstaining from low-quality ones, enabling safer deployment of generative language models.
Traditional training paradigms for extractive and abstractive summarization systems always only use token-level or sentence-level training objectives. However, the output summary is always evaluated from summary-level which leads to the inconsistency in training and evaluation. In this paper, we propose a Contrastive Learning based re-ranking framework for one-stage summarization called COLO. By modeling a contrastive objective, we show that the summarization model is able to directly generate summaries according to the summary-level score without additional modules and parameters. Extensive experiments demonstrate that COLO boosts the extractive and abstractive results of one-stage systems on CNN/DailyMail benchmark to 44.58 and 46.33 ROUGE-1 score while preserving the parameter efficiency and inference efficiency. Compared with state-of-the-art multi-stage systems, we save more than 100 GPU training hours and obtaining 3~8 speed-up ratio during inference while maintaining comparable results.
Current abstractive summarization models either suffer from a lack of clear interpretability or provide incomplete rationales by only highlighting parts of the source document. To this end, we propose the Summarization Program (SP), an interpretable modular framework consisting of an (ordered) list of binary trees, each encoding the step-by-step generative process of an abstractive summary sentence from the source document. A Summarization Program contains one root node per summary sentence, and a distinct tree connects each summary sentence (root node) to the document sentences (leaf nodes) from which it is derived, with the connecting nodes containing intermediate generated sentences. Edges represent different modular operations involved in summarization such as sentence fusion, compression, and paraphrasing. We first propose an efficient best-first search method over neural modules, SP-Search that identifies SPs for human summaries by directly optimizing for ROUGE scores. Next, using these programs as automatic supervision, we propose seq2seq models that generate Summarization Programs, which are then executed to obtain final summaries. We demonstrate that SP-Search effectively represents the generative process behind human summaries using modules that are typically faithful to their intended behavior. We also conduct a simulation study to show that Summarization Programs improve the interpretability of summarization models by allowing humans to better simulate model reasoning. Summarization Programs constitute a promising step toward interpretable and modular abstractive summarization, a complex task previously addressed primarily through blackbox end-to-end neural systems. Supporting code available at https://github.com/swarnaHub/SummarizationPrograms
Social media platforms have become new battlegrounds for anti-social elements, with misinformation being the weapon of choice. Fact-checking organizations try to debunk as many claims as possible while staying true to their journalistic processes but cannot cope with its rapid dissemination. We believe that the solution lies in partial automation of the fact-checking life cycle, saving human time for tasks which require high cognition. We propose a new workflow for efficiently detecting previously fact-checked claims that uses abstractive summarization to generate crisp queries. These queries can then be executed on a general-purpose retrieval system associated with a collection of previously fact-checked claims. We curate an abstractive text summarization dataset comprising noisy claims from Twitter and their gold summaries. It is shown that retrieval performance improves 2x by using popular out-of-the-box summarization models and 3x by fine-tuning them on the accompanying dataset compared to verbatim querying. Our approach achieves Recall@5 and MRR of 35% and 0.3, compared to baseline values of 10% and 0.1, respectively. Our dataset, code, and models are available publicly: https://github.com/varadhbhatnagar/FC-Claim-Det/
Despite the success of recent abstractive summarizers on automatic evaluation metrics, the generated summaries still present factual inconsistencies with the source document. In this paper, we focus on entity-level factual inconsistency, i.e. reducing the mismatched entities between the generated summaries and the source documents. We therefore propose a novel entity-based SpanCopy mechanism, and explore its extension with a Global Relevance component. Experiment results on four summarization datasets show that SpanCopy can effectively improve the entity-level factual consistency with essentially no change in the word-level and entity-level saliency. The code is available at https://github.com/Wendy-Xiao/Entity-based-SpanCopy
Keyphrases are crucial for searching and systematizing scholarly documents. Most current methods for keyphrase extraction are aimed at the extraction of the most significant words in the text. But in practice, the list of keyphrases often includes words that do not appear in the text explicitly. In this case, the list of keyphrases represents an abstractive summary of the source text. In this paper, we experiment with popular transformer-based models for abstractive text summarization using four benchmark datasets for keyphrase extraction. We compare the results obtained with the results of common unsupervised and supervised methods for keyphrase extraction. Our evaluation shows that summarization models are quite effective in generating keyphrases in the terms of the full-match F1-score and BERTScore. However, they produce a lot of words that are absent in the author's list of keyphrases, which makes summarization models ineffective in terms of ROUGE-1. We also investigate several ordering strategies to concatenate target keyphrases. The results showed that the choice of strategy affects the performance of keyphrase generation.
The problems of unfaithful summaries have been widely discussed under the context of abstractive summarization. Though extractive summarization is less prone to the common unfaithfulness issues of abstractive summaries, does that mean extractive is equal to faithful? Turns out that the answer is no. In this work, we define a typology with five types of broad unfaithfulness problems (including and beyond not-entailment) that can appear in extractive summaries, including incorrect coreference, incomplete coreference, incorrect discourse, incomplete discourse, as well as other misleading information. We ask humans to label these problems out of 1600 English summaries produced by 16 diverse extractive systems. We find that 30% of the summaries have at least one of the five issues. To automatically detect these problems, we find that 5 existing faithfulness evaluation metrics for summarization have poor correlations with human judgment. To remedy this, we propose a new metric, ExtEval, that is designed for detecting unfaithful extractive summaries and is shown to have the best performance. We hope our work can increase the awareness of unfaithfulness problems in extractive summarization and help future work to evaluate and resolve these issues. Our data and code are publicly available at https://github.com/ZhangShiyue/extractive_is_not_faithful