Adversarial Examples for Evaluating Reading Comprehension Systems

Jia, Robin; Liang, Percy

doi:10.48550/arxiv.1707.07328

Cited by 186 publications

(157 citation statements)

References 18 publications

Supporting

Mentioning

154

Contrasting

Order By: Relevance

“…Our analysis of the surveyed work shows that NLP models are more robust to adversarial attacks (e.g., word substitution attacks) when trained with robust training (i.e., the IBP) as opposed to normally-trained models which fared poorly (classification accuracy of 36.0%) under the same adversarial attacks [78]. An interesting area of research in this context is whether models trained with data augmentation would be more robust to attacks than robustly trained models.…”

Section: G Insights and Open Directionsmentioning

confidence: 93%

“…We observe that data augmentation, adversarial training, multi-task learning, and robust training all have a positive impact on the classification accuracy of NLP models thereby contributing to robustness. We have also noticed from analyzing the literature that robust training outperforms both adversarial training as well as data augmentation when it comes to robustness to adversarial attacks [78].…”

Section: Insights and Open Directionsmentioning

confidence: 96%

“…MH works by conducting a random walk according to a Markov chain whose stationary distribution is π (the eventual distribution from which the chain will sample). On each step of the MC, a new state is proposed and either accepted or rejected according to a dynamically calculated probability value, called the acceptance criteria [78]. That is, in the long term, the data points from the MC will look similar to the data points from π [60], [59].…”

Section: Breaching Security By Improving Attacksmentioning

confidence: 99%

“…The upper bound can be minimized using the back-propagation approach. The lower bound can be achieved using IBP to measure certified accuracy [78], [122].…”

Section: G Insights and Open Directionsmentioning

confidence: 99%

See 3 more Smart Citations

Robust Natural Language Processing: Recent Advances, Challenges, and Future Directions

Omar¹,

Choi²,

Nyang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Recent natural language processing (NLP) techniques have accomplished high performance on benchmark datasets, primarily due to the significant improvement in the performance of deep learning. The advances in the research community have led to great enhancements in state-of-the-art production systems for NLP tasks, such as virtual assistants, speech recognition, and sentiment analysis. However, such NLP systems still often fail when tested with adversarial attacks. The initial lack of robustness exposed troubling gaps in current models' language understanding capabilities, creating problems when NLP systems are deployed in real life. In this paper, we present a structured overview of NLP robustness research by summarizing the literature in a systemic way across various dimensions. We then take a deep-dive into the various dimensions of robustness, across techniques, metrics, embeddings, and benchmarks. Finally, we argue that robustness should be multi-dimensional, provide insights into current research, identify gaps in the literature to suggest directions worth pursuing to address these gaps.

show abstract

Section: G Insights and Open Directionsmentioning

confidence: 93%

Section: Insights and Open Directionsmentioning

confidence: 96%

Section: Breaching Security By Improving Attacksmentioning

confidence: 99%

“…The upper bound can be minimized using the back-propagation approach. The lower bound can be achieved using IBP to measure certified accuracy [78], [122].…”

Section: G Insights and Open Directionsmentioning

confidence: 99%

See 2 more Smart Citations

Robust Natural Language Processing: Recent Advances, Challenges, and Future Directions

Omar¹,

Choi²,

Nyang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Language modeling is a central problem in natural language processing with wide-ranging applications such as machine translation [2,35,36,40], question answering [7, 20,34,45], article generation [5,30,31], text summarization [22,49], among others. In early work [1,17,21], the dominant approach to train language models is collecting data and training models for specific tasks, which works well under particular data distribution and cases, Title: Angelina Jolie Will Be Visiting Professor at London School of Economics Publish date: 05-24-2016 Domain: www.nytimes.com Body: Angelina Jolie Pitt, the Oscar-winning actress and special envoy for the United Nations refugee agency, has taken on a new role: university professor. Ms. Jolie Pitt will join the London School of Economics' Center for Women, Peace and Security as one of four visiting professors in a new master's program that starts taking applicants in the fall, according to a statement released on Monday by the university, one of Britain's most renowned academic institutions.…”

Section: Introductionmentioning

confidence: 99%

Show and Write: Entity-aware Article Generation with Image Information

Zhang¹,

Gu²,

Plummer³

2021

Preprint

View full text Add to dashboard Cite

Automatically writing long articles is a complex and challenging language generation task. Prior work has primarily focused on generating these articles using humanwritten prompt to provide some topical context and some metadata about the article. That said, for many applications, such as generating news stories, these articles are often paired with images and their captions or alt-text, which in turn are based on real-world events and may reference many different named entities that are difficult to be correctly recognized and predicted by language models. To address these two problems, this paper introduces an Entityaware News Generation method with Image iNformation, ENGIN, to incorporate news image information into language models. ENGIN produces news articles conditioned on both metadata and information such as captions and named entities extracted from images. We also propose an Entity-aware mechanism to help our model better recognize and predict the entity names in news. We perform experiments on two public large-scale news datasets, Good-News and VisualNews. Quantitative results show that our approach improves article perplexity by 4-5 points over the base models. Qualitative results demonstrate the text generated by ENGIN is more consistent with news images. We also perform article quality annotation experiment on the generated articles to validate that our model produces higher-quality articles. Finally, we investigate the effect ENGIN has on methods that automatically detect machinegenerated articles.

show abstract

Conditional BERT Contextual Augmentation

Zang

et al. 2019

Lecture Notes in Computer Science

245

143

View full text Add to dashboard Cite

We propose a novel data augmentation method for labeled sentences called conditional BERT contextual augmentation. Data augmentation methods are often applied to prevent overfitting and improve generalization of deep neural network models. Recently proposed contextual augmentation augments labeled sentences by randomly replacing words with more varied substitutions predicted by language model. BERT demonstrates that a deep bidirectional language model is more powerful than either an unidirectional language model or the shallow concatenation of a forward and backward model. We retrofit BERT to conditional BERT by introducing a new conditional masked language model 1 task. The well trained conditional BERT can be applied to enhance contextual augmentation. Experiments on six various different text classification tasks show that our method can be easily applied to both convolutional or recurrent neural networks classifier to obtain obvious improvement.

show abstract

Adversarial Examples for Evaluating Reading Comprehension Systems

Cited by 186 publications

References 18 publications

Robust Natural Language Processing: Recent Advances, Challenges, and Future Directions

Robust Natural Language Processing: Recent Advances, Challenges, and Future Directions

Show and Write: Entity-aware Article Generation with Image Information

Conditional BERT Contextual Augmentation

Contact Info

Product

Resources

About