Recent work on evaluating grammatical knowledge in pretrained sentence encoders gives a fine-grained view of a small number of phenomena. We introduce a new analysis dataset that also has broad coverage of linguistic phenomena. We annotate the development set of the Corpus of Linguistic Acceptability (CoLA; Warstadt et al., 2018) for the presence of 13 classes of syntactic phenomena including various forms of argument alternations, movement, and modification. We use this analysis set to investigate the grammatical knowledge of three pretrained encoders: BERT (Devlin et al., 2018), GPT (Radford et al., 2018), and the BiLSTM baseline from Warstadt et al. We find that these models have a strong command of complex or non-canonical argument structures like ditransitives (Sue gave Dan a book) and passives (The book was read). Sentences with long distance dependencies like questions (What do you think I ate?) challenge all models, but for these, BERT and GPT have a distinct advantage over the baseline. We conclude that recent sentence encoders, despite showing near-human performance on acceptability classification overall, still fail to make fine-grained grammaticality distinctions for many complex syntactic structures.
New York University（纽约大学）；Facebook AI Research（Facebook人工智能研究院）
This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau e…
University of Salerno（萨勒诺大学）；ICAR-CNR（意大利国家研究委员会信息与高性能计算研究所）；Fondazione Bruno Kessler（布鲁诺·凯斯勒基金会）
The development of automated approaches to linguistic acceptability has been greatly fostered by the availability of the English CoLA corpus, which has also been included in the widely used GLUE benchmark. However, this kind of research for languages other than English, as well as the analysis of cross-lingual approaches, has been hindered by the lack of resources with a comparable size in other languages. We have therefore developed the ItaCoLA corpus, containing almost 10,000 sentences with acceptability judgments, which has been created following the same approach and the same steps as the …
If Probable, Then Acceptable? Understanding Conditional Acceptability Judgments in Large Language Models
MaiNLP, Center for Information and Language Processing, LMU Munich（MaiNLP，信息与语言处理中心，慕尼黑大学）；Munich Center for Machine Learning (MCML), Munich（慕尼黑机器学习中心（MCML），慕尼黑）
Linguistic Acceptability is the task of determining whether a sentence is grammatical or ungrammatical. It has applications in several use cases like Question-Answering, Natural Language Generation, Neural Machine Translation, where grammatical correctness is crucial. In this paper we aim to understand the decision-making process of BERT (Devlin et al., 2019) in distinguishing between Linguistically Acceptable sentences (LA) and Linguistically Unacceptable sentences (LUA). We leverage Layer Integrated Gradients Attribution Scores (LIG) to explain the Linguistic Acceptability criteria that are …
RuCoLA: Russian Corpus of Linguistic Acceptability
SberDevices；ABBYY；HSE University（高等经济大学）；Yandex；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）；Center for Information and Language Processing (CIS), MaiNLP lab, LMU Munich（慕尼黑大学信息与语言处理中心（CIS）MaiNLP实验室）
Knowledge of syntax includes knowledge of rare, idiosyncratic constructions. LLMs must overcome frequency biases in order to master such constructions. In this study, I prompt GPT-3 to give acceptability judgments on the English-language Article + Adjective + Numeral + Noun construction (e.g., "a lovely five days"). I validate the prompt using the CoLA corpus of acceptability judgments and then zero in on the AANN construction. I compare GPT- 3's judgments to crowdsourced human judgments on a subset of sentences. GPT-3's judgments are broadly similar to human judgments and generally align with…
Beijing International Studies University（北京第二外国语学院）；Penn State（宾夕法尼亚州立大学）；Google Research（谷歌研究院）
Native speakers can judge whether a sentence is an acceptable instance of their language. Acceptability provides a means of evaluating whether computational language models are processing language in a human-like manner. We test the ability of computational language models, simple language features, and word embeddings to predict native English speakers judgments of acceptability on English-language essays written by non-native speakers. We find that much of the sentence acceptability variance can be captured by a combination of features including misspellings, word order, and word similarity …
We present DaLAJ 1.0, a Dataset for Linguistic Acceptability Judgments for Swedish, comprising 9 596 sentences in its first version; and the initial experiment using it for the binary classification task. DaLAJ is based on the SweLL second language learner data, consisting of essays at different levels of proficiency. To make sure the dataset can be freely available despite the GDPR regulations, we have sentence-scrambled learner essays and removed part of the metadata about learners, keeping for each sentence only information about the mother tongue and the level of the course where the essay…
What Makes it Ok to Set a Fire? Iterative Self-distillation of Contexts and Rationales for Disambiguating Defeasible Social and Moral Situations
Paul G. Allen School of Computer Science & Engineering, University of Washington（华盛顿大学保罗·G·艾伦计算机科学与工程学院）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）
Shanghai Jiao Tong University（上海交通大学）；City University of Hong Kong（香港城市大学）；University of Pennsylvania（宾夕法尼亚大学）；Brigham Young University（杨百翰大学）；University of Washington（华盛顿大学）；Indiana University Bloomington（印第安纳大学伯明顿分校）
Pretrained transformer-based language models have achieved state of the art across countless tasks in natural language processing. These models are highly expressive, comprising at least a hundred million parameters and a dozen layers. Recent evidence suggests that only a few of the final layers need to be fine-tuned for high quality on downstream tasks. Naturally, a subsequent research question is, "how many of the last layers do we need to fine-tune?" In this paper, we precisely answer this question. We examine two recent pretrained language models, BERT and RoBERTa, across standard tasks in…
ETH Zürich（苏黎世联邦理工学院）；University of Cambridge（剑桥大学）；University of Zurich（苏黎世大学）；University of Potsdam（波茨坦大学）；Massachusetts Institute of Technology（麻省理工学院）
The uniform information density (UID) hypothesis posits a preference among language users for utterances structured such that information is distributed uniformly across a signal. While its implications on language production have been well explored, the hypothesis potentially makes predictions about language comprehension and linguistic acceptability as well. Further, it is unclear how uniformity in a linguistic signal -- or lack thereof -- should be measured, and over which linguistic unit, e.g., the sentence or language level, this uniformity should hold. Here we investigate these facets of…
Acceptability Judgements via Examining the Topology of Attention Maps
Skolkovo Institute of Science and Technology（斯科尔科沃科学技术研究院）；AIRI（AIRI（人工智能研究院））；SberDevices（SberDevices（俄联邦储蓄银行设备公司））；HSE University（高等经济大学）；Huawei Noah’s Ark lab（华为诺亚方舟实验室）；Center for Information and Language Processing (CIS), LMU Munich（慕尼黑大学信息与语言处理中心）；CNRS（法国国家科学研究中心）
本文利用拓扑数据分析提取注意力图的几何特征，用于语法可接受性二元判断和最小对比较，在多语言 CoLA 与 BLiMP 上显著提升性能并达到人类水平。
As natural language processing methods are increasingly deployed in real-world scenarios such as healthcare, legal systems, and social science, it becomes necessary to recognize the role they potentially play in shaping social biases and stereotypes. Previous work has revealed the presence of social biases in widely used word embeddings involving gender, race, religion, and other social constructs. While some methods were proposed to debias these word-level embeddings, there is a need to perform debiasing at the sentence-level given the recent shift towards new contextualized sentence represen…
Motivated by recent findings on the probabilistic modeling of acceptability judgments, we propose syntactic log-odds ratio (SLOR), a normalized language model score, as a metric for referenceless fluency evaluation of natural language generation output at the sentence level. We further introduce WPSLOR, a novel WordPiece-based version, which harnesses a more compact language model. Even though word-overlap metrics like ROUGE are computed with the help of hand-written references, our referenceless methods obtain a significantly higher correlation with human fluency scores on a benchmark dataset…
To understand what kinds of linguistic knowledge are encoded by pretrained Chinese language models (LMs), we introduce the benchmark of Sino LINGuistics (SLING), which consists of 38K minimal sentence pairs in Mandarin Chinese grouped into 9 high-level linguistic phenomena. Each pair demonstrates the acceptability contrast of a specific syntactic or semantic phenomenon (e.g., The keys are lost vs. The keys is lost), and an LM should assign lower perplexity to the acceptable sentence. In contrast to the CLiMP dataset (Xiang et al., 2021), which also contains Chinese minimal pairs and was create…
University of Toronto（多伦多大学）；Vector Institute for Artificial Intelligence（向量人工智能研究所）；Unity Health Toronto（多伦多联合健康）
Automatically evaluating text-based, non-task-oriented dialogue systems (i.e., `chatbots') remains an open problem. Previous approaches have suffered challenges ranging from poor correlation with human judgment to poor generalization and have often required a gold standard reference for comparison or human-annotated data. Extending existing evaluation methods, we propose that a metric based on linguistic features may be able to maintain good correlation with human judgment and be interpretable, without requiring a gold-standard reference or human-annotated data. To support this proposition, we…
Automatic phylogenetic inference plays an increasingly important role in computational historical linguistics. Most pertinent work is currently based on expert cognate judgments. This limits the scope of this approach to a small number of well-studied language families. We used machine learning techniques to compile data suitable for phylogenetic inference from the ASJP database, a collection of almost 7,000 phonetically transcribed word lists over 40 concepts, covering two third of the extant world-wide linguistic diversity. First, we estimated Pointwise Mutual Information scores between soun…
Language models align with human judgments on key grammatical constructions
Kempner Institute for the Study of Natural and Artificial Intelligence, Harvard University（哈佛大学坎普ner自然与人工智能研究所）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）；University of Wisconsin-Madison（威斯康星大学麦迪逊分校）；Georgia Tech（佐治亚理工学院）；Massachusetts Institute of Technology（麻省理工学院）
NLP is currently dominated by general-purpose pretrained language models like RoBERTa, which achieve strong performance on NLU tasks through pretraining on billions of words. But what exact knowledge or skills do Transformer LMs learn from large-scale pretraining that they cannot learn from less data? We adopt four probing methods---classifier probing, information-theoretic probing, unsupervised relative acceptability judgment, and fine-tuning on NLU tasks---and draw learning curves that track the growth of these different measures of linguistic ability with respect to pretraining data volume …
Peking University（北京大学）；Shanghai Jiao Tong University（上海交通大学）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）
We conduct a thorough study to diagnose the behaviors of pre-trained language encoders (ELMo, BERT, and RoBERTa) when confronted with natural grammatical errors. Specifically, we collect real grammatical errors from non-native speakers and conduct adversarial attacks to simulate these errors on clean text data. We use this approach to facilitate debugging models on downstream applications. Results confirm that the performance of all tested models is affected but the degree of impact varies. To interpret model behaviors, we further design a linguistic acceptability task to reveal their abilitie…
Department of Computer, Control and Management Engineering, Sapienza University of Rome（计算机、控制与管理工程系，罗马萨皮恩扎大学）；Department of Computer Science, Sapienza University of Rome（计算机科学系，罗马萨皮恩扎大学）；Department of Legal, Social, and Educational Sciences, Tuscia University（法律、社会与教育科学系，图斯西亚大学）；Department of Psychology, Sapienza University of Rome（心理学系，罗马萨皮恩扎大学）
Epistemological Fault Lines Between Human and Artificial Intelligence
Department of Computer Science, Sapienza University of Rome, Rome, Italy；Department of Psychology, University of Milan Bicocca, Milan, Italy；Faculty of Natural Sciences；Mathematics, University of Maribor, Maribor, Slovenia；Community Healthcare Center Dr. Adolf Drolc Maribor, Maribor, Slovenia；University College, Korea University, Seoul, Republic of Korea；Department of Physics, Kyung Hee University, Seoul, Republic of Korea
