An adversarial example is an input transformed by small perturbations that machine learning models consistently misclassify. While there are a number of methods proposed to generate adversarial examples for text data, it is not trivial to assess the quality of these adversarial examples, as minor perturbations (such as changing a word in a sentence) can lead to a significant shift in their meaning, readability and classification label. In this paper, we propose an evaluation framework consisting of a set of automatic evaluation metrics and human evaluation guidelines, to rigorously assess the quality of adversarial examples based on the aforementioned properties. We experiment with six benchmark attacking methods and found that some methods generate adversarial examples with poor readability and content preservation. We also learned that multiple factors could influence the attacking performance, such as the length of the text inputs and architecture of the classifiers.
The Elephant in the Room: Analyzing the Presence of Big Tech in Natural Language Processing Research
Institute for Better Health（健康促进研究所）；University of Göttingen（哥廷根大学）；Université Paris-Saclay（巴黎萨克雷大学）；Sorbonne Université（索邦大学）；LORIA（洛里亚实验室）；National Research Council（国家研究委员会）
Assessing the Applicability of Natural Language Processing to Traditional Social Science Methodology: A Case Study in Identifying Strategic Signaling Patterns in Presidential Directives
ADAPT Centre（ADAPT中心）；Dublin City University（都柏林城市大学）
Reproducibility has become an intensely debated topic in NLP and ML over recent years, but no commonly accepted way of assessing reproducibility, let alone quantifying it, has so far emerged. The assumption has been that wider scientific reproducibility terminology and definitions are not applicable to NLP/ML, with the result that many different terms and definitions have been proposed, some diametrically opposed. In this paper, we test this assumption, by taking the standard terminology and definitions from metrology and applying them directly to NLP/ML. We find that we are able to straightfo…
Missing Information, Unresponsive Authors, Experimental Flaws: The Impossibility of Assessing the Reproducibility of Previous Human Evaluations in NLP
ADAPT/DCU（ADAPT中心/都柏林城市大学）；University of Aberdeen（阿伯丁大学）；Bielefeld University（比勒费尔德大学）；Dataminr（Dataminr公司）；Edinburgh Napier University（爱丁堡龙比亚大学）；Georgia Tech（佐治亚理工学院）；Google Research（谷歌研究院）；Groningen University（格罗宁根大学）；Heriot-Watt University（赫瑞-瓦特大学）；Charles University Prague（布拉格查理大学）；Tohoku University（东北大学）；Peking University（北京大学）；Technological University Dublin（都柏林理工大学）；Tilburg University（蒂尔堡大学）；trivago（trivago公司）；University of Applied Sciences Darmstadt（达姆施塔特应用科学大学）；University of Illinois Chicago（伊利诺伊大学芝加哥分校）；Universidade de Santiago de Compostela（圣地亚哥德孔波斯特拉大学）；Universitat Pompeu Fabra（庞培法布拉大学）；Utrecht University（乌得勒支大学）；Zurich University of Applied Sciences（苏黎世应用科学大学）
Technion – Israel Institute of Technology（以色列理工学院）；Stanford University（斯坦福大学）；Ghent University（根特大学）
The increasing size and complexity of modern ML systems has improved their predictive capabilities but made their behavior harder to explain. Many techniques for model explanation have been developed in response, but we lack clear criteria for assessing these techniques. In this paper, we cast model explanation as the causal inference problem of estimating causal effects of real-world concepts on the output behavior of ML models given actual input data. We introduce CEBaB, a new benchmark dataset for assessing concept-based explanation methods in Natural Language Processing (NLP). CEBaB consis…
Performance of diverse evaluation metrics in NLP-based assessment and text generation of consumer complaints
School of Mathematical and Statistical Sciences, Western University（数学与统计科学学院，西部大学）；Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai（流行病学与政策部门，辛辛那提医学院）；Institute for Health Care Delivery Science, Icahn School of Medicine at Mount Sinai（健康护理科学研究所，辛辛那提医学院）
Assessing Thai Dialect Performance in LLMs with Automatic Benchmarks and Human Evaluation
AI Singapore（新加坡人工智能中心）；SCB 10X；School of Information Science and Technology, VISTEC（VISTEC信息科学与技术学院）；Chulalongkorn University（朱拉隆功大学）；National University of Singapore（新加坡国立大学）
While self-supervised learning has made rapid advances in natural language processing, it remains unclear when researchers should engage in resource-intensive domain-specific pretraining (domain pretraining). The law, puzzlingly, has yielded few documented instances of substantial gains to domain pretraining in spite of the fact that legal language is widely seen to be unique. We hypothesize that these existing results stem from the fact that existing legal NLP tasks are too easy and fail to meet conditions for when domain pretraining can help. To address this, we first present CaseHOLD (Case …
Recent advances in NLP have been attributed to the emergence of large-scale pre-trained language models. GPT-2, in particular, is suited for generation tasks given its left-to-right language modeling objective, yet the linguistic quality of its generated text has largely remain unexplored. Our work takes a step in understanding GPT-2's outputs in terms of discourse coherence. We perform a comprehensive study on the validity of explicit discourse relations in GPT-2's outputs under both organic generation and fine-tuned scenarios. Results show GPT-2 does not always generate text containing valid…
Hasso Plattner Institute（哈索·普拉特纳研究所）；Rutgers University（罗格斯大学）；University of Potsdam（波茨坦大学）
Emojis have become ubiquitous in digital communication, due to their visual appeal as well as their ability to vividly convey human emotion, among other factors. The growing prominence of emojis in social media and other instant messaging also leads to an increased need for systems and tools to operate on text containing emojis. In this study, we assess this support by considering test sets of tweets with emojis, based on which we perform a series of experiments investigating the ability of prominent NLP and text processing tools to adequately process them. In particular, we consider tokenizat…
University Of South Florida（南佛罗里达大学）；Northern Illinois University（北伊利诺伊大学）；Ajivar LLC（Ajivar有限责任公司）
In this paper, we report experimental results on assessing the impact of COVID-19 on college students by processing free-form texts generated by them. By free-form texts, we mean textual entries posted by college students (enrolled in a four year US college) via an app specifically designed to assess and improve their mental health. Using a dataset comprising of more than 9000 textual entries from 1451 students collected over four months (split between pre and post COVID-19), and established NLP techniques, a) we assess how topics of most interest to student change between pre and post COVID-1…
Neural network NLP models are vulnerable to small modifications of the input that maintain the original meaning but result in a different prediction. In this paper, we focus on robustness of text classification against word substitutions, aiming to provide guarantees that the model prediction does not change if a word is replaced with a plausible alternative, such as a synonym. As a measure of robustness, we adopt the notion of the maximal safe radius for a given input text, which is the minimum distance in the embedding space to the decision boundary. Since computing the exact maximal safe ra…
TCM-3CEval: A Triaxial Benchmark for Assessing Responses from Large Language Models in Traditional Chinese Medicine
School of Acupuncture-Moxibustion and Tuina, Shanghai University of Traditional Chinese Medicine（上海中医药大学针灸推拿学院）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；China Academy of Chinese Medical Sciences（中国中医科学院）
Indiana University（印第安纳大学）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）；University of Pennsylvania（宾夕法尼亚大学）
Empirical research in Natural Language Processing (NLP) has adopted a narrow set of principles for assessing hypotheses, relying mainly on p-value computation, which suffers from several known issues. While alternative proposals have been well-debated and adopted in other fields, they remain rarely discussed or used within the NLP community. We address this gap by contrasting various hypothesis assessment techniques, especially those not commonly used in the field (such as evaluations based on Bayesian inference). Since these statistical techniques differ in the hypotheses they can support, we…
Information School, the University of Sheffield（谢菲尔德大学信息学院）；Vamstar Ltd.（Vamstar有限公司）
While text mining and NLP research has been established for decades, there remain gaps in the literature that reports the use of these techniques in building real-world applications. For example, they typically look at single and sometimes simplified tasks, and do not discuss in-depth data heterogeneity and inconsistency that is common in real-world problems or their implication on the development of their methods. Also, few prior work has focused on the healthcare domain. In this work, we describe an industry project that developed text mining and NLP solutions to mine millions of heterogeneo…
NLP systems use language models such as Masked Language Models (MLMs) that are pre-trained on large quantities of text such as Wikipedia create representations of language. BERT is a powerful and flexible general-purpose MLM system developed using unlabeled text. Pre-training on large quantities of text also has the potential to transparently embed the cultural and social biases found in the source text into the MLM system. This study aims to compare biases in general purpose and medical MLMs with the StereoSet bias assessment tool. The general purpose MLMs showed significant bias overall, wit…
Assessing socio-economic climate impacts from text data
Helmholtz Centre for Environmental Research, Germany（德国海德堡环境研究中心）；Leipzig University, Germany（德国莱比锡大学）；UCLouvain Brussels, Belgium（比利时布鲁塞尔UCLouvain）；Ecole des Ponts, France（法国École des Ponts）；Université PSL, École Polytechnique, Institut Polytechnique de Paris, Sorbonne Université, CNRS, France（法国巴黎理工学院、巴黎理工研究院、巴黎政治学院、法国国家科学研究中心）；RISE Research Institutes of Sweden, Sweden（瑞典RISE研究机构）；Swedish Centre for Impacts of Climate Extremes (Climes), Sweden（瑞典气候极端影响研究中心）；Vrije Universiteit Brussel, Belgium（比利时布鲁塞尔自由大学）；TUD Dresden University of Technology, Germany（德国德累斯顿理工大学）；Uppsala University, Sweden（瑞典乌普萨拉大学）；Politecnico di Milano, Italy（意大利米兰理工大学）；Istituto Universitario di Studi Superiori IUSS Pavia, Italy（意大利皮亚维亚大学研究院）；Center for Climate Change and Transformation, Eurac Research, Italy（意大利Eurac研究机构气候变化与转型中心）；Royal Museum for Central Africa, Belgium（比利时皇家中非博物馆）；Technopolis Group, Germany（德国Technopolis集团）
Exploring the Use of Large Language Models for Reference-Free Text Quality Evaluation: An Empirical Study
Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；Peng Cheng Laboratory（鹏城实验室）；Guangdong Provincial Key Laboratory of Novel Security Intelligence Technologies（广东省新型安全智能技术重点实验室）
