总结 | ACL2022主会论文分类整理(一）

大家好，我是对白。

ACL 2022是CCF A类会议，人工智能领域自然语言处理（Natural Language Processing，NLP）方向最权威的国际会议之一。第60届计算语言学协会计划于今年5月22日-5月27日在爱尔兰都柏林召开。

本文对ACL 2022接受列表中的的602篇主会长文论文，**按不同的研究主题进行分类整理（分类标准参考 ACL 官方投稿主题），**整理过程中难免有疏漏，欢迎大家在下方评论留言，交流探讨！

论文列表已经同步更新到 GitHub，欢迎大家关注和 Star。

目录

• Adversarial attack and Robustness【对抗攻击和鲁棒性】

• Dialogue and Interactive Systems【对话与交互系统】

• Discourse and Pragmatics【语篇和语用学】

• Data Augmentation【数据增广】

• Generation【文本生成】

• Information Extraction【信息抽取】

• Information Retrieval and Text Mining【信息检索与文本挖掘】

• Interpretability and Analysis of Models for NLP【NLP模型的可解释性与分析】

• Language Model【语言模型】

• Machine Learning for NLP【NLP中的机器学习】

• Machine Translation and Multilinguality【机器翻译与多语】

• Question Answering【问答与理解】

• Resources and Evaluation【数据集与评估方法】

• Sentence-level Semantics, Textual Classification, and Other Areas【句子级语义和关系推理】

• Semantics and Syntax Parsing【语义与句法解析】

• Speech and Multimodality【语音与多模态】

• Summation【摘要】

• Knowledge Graph【知识图谱】

• Special Track【特殊任务】

Adversarial attack and Robustness【对抗攻击和鲁棒性】

• Adversarial Authorship Attribution for Deobfuscation

• Adversarial Soft Prompt Tuning for Cross-Domain Sentiment Analysis

• Flooding-X: Improving BERT’s Resistance to Adversarial Attacks via LossRestricted Fine-Tuning

• From the Detection of Toxic Spans in Online Discussions to the Analysis of Toxic-to-Civil Transfer

• Imputing Out-of-Vocabulary Embeddings with LOVE Makes Language Models Robust with Little Cost

• ParaDetox: Detoxification with Parallel Data

• Pass off Fish Eyes for Pearls: Attacking Model Selection of Pre-trained Models

• SHIELD: Defending Textual Neural Networks against Multiple Black-Box

• Adversarial Attacks with Stochastic Multi-Expert Patcher

• Towards Robustness of Text-to-SQL Models Against Natural and Realistic Adversarial Table Perturbation

• ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection

Dialogue and Interactive Systems【对话与交互系统】

• A Model-agnostic Data Manipulation Method for Persona-based Dialogue Generation

• A Taxonomy of Empathetic Questions in Social Dialogs

• Achieving Conversational Goals with Unsupervised Post-hoc Knowledge Injection

• Achieving Reliable Human Assessment of Open-Domain Dialogue Systems

• An Interpretable Neuro-Symbolic Reasoning Framework for Task-Oriented Dialogue Generation

• Beyond Goldfish Memory: Long-Term Open-Domain Conversation

• Beyond the Granularity: Multi-Perspective Dialogue Collaborative Selection for Dialogue State Tracking

• CASPI Causal-aware Safe Policy Improvement for Task-oriented Dialogue

• ChatMatch: Evaluating Chatbots by Autonomous Chat Tournaments

• CICERO: A Dataset for Contextualized Commonsense Inference in Dialogues

• Contextual Fine-to-Coarse Distillation for Coarse-grained Response Selection in Open-Domain Conversations

• Continual Prompt Tuning for Dialog State Tracking

• DEAM: Dialogue Coherence Evaluation using AMR-based Semantic Manipulations

• DialogVED: A Pre-trained Latent Variable Encoder-Decoder Model for Dialog Response Generation

• Dynamic Schema Graph Fusion Network for Multi-Domain Dialogue State Tracking

• GlobalWoZ: Globalizing MultiWoZ to Develop Multilingual Task-Oriented Dialogue Systems

• HeterMPC: A Heterogeneous Graph Neural Network for Response Generation in Multi-Party Conversations

• Improving Multi-label Malevolence Detection in Dialogues through Multifaceted Label Correlation Enhancement

• Interactive Word Completion for Plains Cree

• Internet-Augmented Dialogue Generation

• Knowledge Enhanced Reflection Generation for Counseling Dialogues

• M3ED: Multi-modal Multi-scene Multi-label Emotional Dialogue Database

• MISC: A Mixed Strategy-Aware Model integrating COMET for Emotional Support Conversation

• Multi-Party Empathetic Dialogue Generation: A New Task for Dialog Systems

• Multi-Task Pre-Training for Plug-and-Play Task-Oriented Dialogue System

• Multimodal Dialogue Response Generation

• Online Semantic Parsing for Latency Reduction in Task-Oriented Dialogue

• Other Roles Matter! Enhancing Role-Oriented Dialogue Summarization via Role Interactions

• ProphetChat: Enhancing Dialogue Generation with Simulation of Future Conversation

• QAConv: Question Answering on Informative Conversations

• SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures

• SafetyKit: First Aid for Measuring Safety in Open-domain Conversational Systems

• SalesBot: Transitioning from Chit-Chat to Task-Oriented Dialogues

• Should a Chatbot be Sarcastic? Understanding User Preferences Towards Sarcasm Generation

• Situated Dialogue Learning through Procedural Environment Generation

• Structural Characterization for Dialogue Disentanglement

• The AI Doctor Is In: A Survey of Task-Oriented Dialogue Systems for Healthcare Applications

• There Are a Thousand Hamlets in a Thousand People’s Eyes: Enhancing Knowledge-grounded Dialogue with Personal Memory

• Think Before You Speak: Explicitly Generating Implicit Commonsense Knowledge for Response Generation

• UniTranSeR: A Unified Transformer Semantic Representation Framework for Multimodal Task-Oriented Dialog System

• What does the sea say to the shore? A BERT based DST style approach for speaker to dialogue attribution in novels

• Where to Go for the Holidays: Towards Mixed-Type Dialogs for Clarification of User Goals

• Speaker Information Can Guide Models to Better Inductive Biases: A Case Study On Predicting Code-Switching

Discourse and Pragmatics【语篇和语用学】

• CoCoLM: Complex Commonsense Enhanced Language Model with Discourse Relations

• Context Matters: A Pragmatic Study of PLMs’ Negation Understanding

• Learning to Mediate Disparities Towards Pragmatic Communication

• Modeling Persuasive Discourse to Adaptively Support Students’ Argumentative Writing

• Neural reality of argument structure constructions

• Probing for Predicate Argument Structures in Pretrained Language Models

• RST Discourse Parsing with Second-Stage EDU-Level Pre-training

Data Augmentation【数据增广】

• An Investigation of the (In)effectiveness of Counterfactually Augmented Data

• CipherDAug: Ciphertext based Data Augmentation for Neural Machine Translation

• Continual Few-shot Relation Learning via Embedding Space Regularization and Data Augmentation

• Deduplicating Training Data Makes Language Models Better

• FlipDA: Effective and Robust Data Augmentation for Few-Shot Learning

• Generating Data to Mitigate Spurious Correlations in Natural Language Inference Datasets

• Keywords and Instances: A Hierarchical Contrastive Learning Framework Unifying Hybrid Granularities for Text Generation

• MELM: Data Augmentation with Masked Entity Language Modeling for LowResource NER

• PromDA: Prompt-based Data Augmentation for Low-Resource NLU Tasks

• Synthetic Question Value Estimation for Domain Adaptation of Question Answering

• Training Data is More Valuable than You Think: A Simple and Effective Method by Retrieving from Training Data

Generation【文本生成】

• A Token-level Reference-free Hallucination Detection Benchmark for Freeform Text Generation

• A Well-Composed Text is Half Done! Composition Sampling for Diverse Conditional Generation

• Accurate Online Posterior Alignments for Principled Lexically-Constrained Decoding

• Active Evaluation: Efficient NLG Evaluation with Few Pairwise Comparisons

• AraT5: Text-to-Text Transformers for Arabic Language Generation

• Continual Sequence Generation with Adaptive Compositional Modules

• Controllable Dictionary Example Generation: Generating Example Sentences for Specific Targeted Audiences

• CTRLEval: An Unsupervised Reference-Free Metric for Evaluating Controlled Text Generation

• Few-shot Controllable Style Transfer for Low-Resource Multilingual Settings

• Fine-Grained Controllable Text Generation Using Non-Residual Prompting

• Flexible Generation from Fragmentary Linguistic Input

• FrugalScore: Learning Cheaper, Lighter and Faster Evaluation Metrics for Automatic Text Generation

• Generating Scientific Definitions with Controllable Complexity

• Hierarchical Sketch Induction for Paraphrase Generation

• How Do Seq2Seq Models Perform on End-to-End Data-to-Text Generation?

• Hybrid Semantics for Goal-Directed Natural Language Generation

• Improving Compositional Generalization with Self-Training for Data-to-Text Generation

• Improving Personalized Explanation Generation through Visualization

• Inducing Positive Perspectives with Text Reframing

• latent-GLAT: Glancing at Latent Variables for Parallel Text Generation

• Lexical Knowledge Internalization for Neural Dialog Generation

• Mix and Match: Learning-free Controllable Text Generationusing Energy Language Models

• Multitasking Framework for Unsupervised Simple Definition Generation

• Neural Pipeline for Zero-Shot Data-to-Text Generation

• Non-neural Models Matter: a Re-evaluation of Neural Referring Expression Generation Systems

• ODE Transformer: An Ordinary Differential Equation-Inspired Model for Sequence Generation

• Overlap-based Vocabulary Generation Improves Cross-lingual Transfer Among Related Languages

• PLANET: Dynamic Content Planning in Autoregressive Transformers for Long-form Text Generation

• Predicate-Argument Based Bi-Encoder for Paraphrase Identification

• Principled Paraphrase Generation with Parallel Corpora

• Quality Controlled Paraphrase Generation

• Rare Tokens Degenerate All Tokens: Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings

• RoMe: A Robust Metric for Evaluating Natural Language Generation

• Semi-Supervised Formality Style Transfer with Consistency Training

• So Different Yet So Alike! Constrained Unsupervised Text Style Transfer

• Spurious Correlations in Reference-Free Evaluation of Text Generation

• Tailor: Generating and Perturbing Text with Semantic Controls

• Towards Better Characterization of Paraphrases

• Uncertainty Determines the Adequacy of the Mode and the Tractability of Decoding in Sequence-to-Sequence Models

• An Imitation Learning Curriculum for Text Editing with Non-Autoregressive Models

• Understanding Iterative Revision from Human-Written Text

Information Extraction【信息抽取】

• Alignment-Augmented Consistent Translation for Multilingual Open Information Extraction

• Automatic Error Analysis for Document-level Information Extraction

• BenchIE: A Framework for Multi-Faceted Fact-Based Open Information Extraction Evaluation

• Dynamic Global Memory for Document-level Argument Extraction

• Dynamic Prefix-Tuning for Generative Template-based Event Extraction

• FaVIQ: FAct Verification from Information-seeking Questions

• FormNet: Structural Encoding beyond Sequential Modeling in Form Document Information Extraction

• Generating Scientific Claims for Zero-Shot Scientific Fact Checking

• JointCL: A Joint Contrastive Learning Framework for Zero-Shot Stance Detection

• KNN-Contrastive Learning for Out-of-Domain Intent Classification

• Legal Judgment Prediction via Event Extraction with Constraints

• MILIE: Modular & Iterative Multilingual Open Information Extraction

• Modeling U.S. State-Level Policies by Extracting Winners and Losers from Legislative Texts

• OIE@OIA: an Adaptable and Efficient Open Information Extraction Framework

• Packed Levitated Marker for Entity and Relation Extraction

• Pre-training to Match for Unified Low-shot Relation Extraction

• Prompt for Extraction? PAIE: Prompting Argument Interaction for Event Argument Extraction

• Retrieval-guided Counterfactual Generation for QA

• Right for the Right Reason: Evidence Extraction for Trustworthy Tabular Reasoning

• Saliency as Evidence: Event Detection with Trigger Saliency Attribution

• Text-to-Table: A New Way of Information Extraction

• Toward Interpretable Semantic Textual Similarity via Optimal Transportbased Contrastive Sentence Learning

• Transkimmer: Transformer Learns to Layer-wise Skim

• Unified Structure Generation for Universal Information Extraction

Information Retrieval and Text Mining【信息检索与文本挖掘】

• Automatic Identification and Classification of Bragging in Social Media

• Bilingual alignment transfers to multilingual alignment for unsupervised parallel text mining

• Can Unsupervised Knowledge Transfer from Social Discussions Help Argument Mining?

• ClarET: Pre-training a Correlation-Aware Context-To-Event Transformer for Event-Centric Generation and Classification

• Cross-Lingual Phrase Retrieval

• Learning to Rank Visual Stories From Human Ranking Data

• Multi-View Document Representation Learning for Open-Domain Dense Retrieval

• New Intent Discovery with Pre-training and Contrastive Learning

• Pre-training and Fine-tuning Neural Topic Model: A Simple yet Effective Approach to Incorporating External Knowledge

• RELiC: Retrieving Evidence for Literary Claims

• Retrieval-guided Counterfactual Generation for QA

• SDR: Efficient Neural Re-ranking using Succinct Document Representation

• Sentence-aware Contrastive Learning for Open-Domain Passage Retrieval

• Show Me More Details: Discovering Hierarchies of Procedures from Semistructured Web Data

• Training Data is More Valuable than You Think: A Simple and Effective Method by Retrieving from Training Data

• UCTopic: Unsupervised Contrastive Learning for Phrase Representations and Topic Mining

• Unsupervised Corpus Aware Language Model Pre-training for Dense Passage Retrieval

• Zoom Out and Observe: News Environment Perception for Fake News Detection

Interpretability and Analysis of Models for NLP【NLP模型的可解释性与分析】

• A Closer Look at How Fine-tuning Changes BERT

• A Comparative Study of Faithfulness Metrics for Model Interpretability Methods

• A Comparison of Strategies for Source-Free Domain Adaptation

• Active Evaluation: Efficient NLG Evaluation with Few Pairwise Comparisons

• Adaptive Testing and Debugging of NLP Models

• An Empirical Study of Memorization in NLP

• An Empirical Study on Explanations in Out-of-Domain Settings

• An Empirical Survey of the Effectiveness of Debiasing Techniques for Pretrained Language Models

• An Investigation of the (In)effectiveness of Counterfactually Augmented Data

• Can Explanations Be Useful for Calibrating Black Box Models?

• Can Pre-trained Language Models Interpret Similes as Smart as Human?

• Can Prompt Probe Pretrained Language Models? Understanding the Invisible Risks from a Causal View

• Can Synthetic Translations Improve Bitext Quality?

• Can Transformer be Too Compositional? Analysing Idiom Processing in Neural Machine Translation

• Causal Probing for Grammatical Number: From Encoding to Usage

• Coherence boosting: When your pretrained language model is not paying enough attention

• Context Matters: A Pragmatic Study of PLMs’ Negation Understanding

• Cross-Lingual Ability of Multilingual Masked Language Models: A Study of Language Structure

• Dataset Geography: Mapping Language Data to Language Users

• Do Transformer Models Show Similar Attention Patterns to Task-Specific Human Gaze?

• Does Recommend-Revise Produce Reliable Annotations? An Analysis on Missing Instances in DocRED

• Explanation Graph Generation via Pre-trained Language Models: An Empirical Study with Contrastive Learning

• Finding Structural Knowledge in Multimodal-BERT

• Generating Biographies on Wikipedia: The Impact of Gender Bias on the Retrieval-Based Generation of Women Biographies

• GPT-D: Inducing Dementia-related Linguistic Anomalies by Deliberate Degradation of Artificial Neural Language Models

• How can NLP Help Revitalize Endangered Languages? A Case Study and Roadmap for the Cherokee Language

• ILDAE: Instance-Level Difficulty Analysis of Evaluation Data

• IMPLI: Investigating NLI Models’ Performance on Figurative Language

• Improving Generalizability in Implicitly Abusive Language Detection with Concept Activation Vectors

• Interpretability for Language Learners Using Example-Based Grammatical Error Correction

• Interpreting Character Embeddings With Perceptual Representations: The Case of Shape, Sound, and Color

• Investigating Failures of Automatic Translation in the Case of Unambiguous Gender

• Investigating Non-local Features for Neural Constituency Parsing

• Is Attention Explanation? An Introduction to the Debate

• Life after BERT: What do Other Muppets Understand about Language?

• Low-Rank Softmax Can Have Unargmaxable Classes in Theory but Rarely in Practice

• Measuring Fairness of Text Classifiers via Prediction Sensitivity

• Memorisation versus Generalisation in Pre-trained Language Models

• Metaphors in Pre-Trained Language Models: Probing and Generalization Across Datasets and Languages

• On the Sensitivity and Stability of Model Interpretations in NLP

• Pretraining with Artificial Language: Studying Transferable Knowledge in Language Models

• Probing as Quantifying Inductive Bias

• Probing Simile Knowledge from Pre-trained Language Models

• ProtoTEx: Explaining Model Decisions with Prototype Tensors

• Reports of personal experiences and stories in argumentation: datasets and analysis

• Rewire-then-Probe: A Contrastive Recipe for Probing Biomedical Knowledge of Pre-trained Language Models

• Sense Embeddings are also Biased – Evaluating Social Biases in Static and Contextualised Sense Embeddings

• Signal in Noise: Exploring Meaning Encoded in Random Character Sequences with Character-Aware Language Models

• Systematic Inequalities in Language Technology Performance across the World’s Languages

• That Is a Suspicious Reaction!: Interpreting Logits Variation to Detect NLP Adversarial Attacks

• The Dangers of Underclaiming: Reasons for Caution When Reporting How NLP Systems Fail

• The Moral Debater: A Study on the Computational Generation of Morally Framed Arguments

• The Paradox of the Compositionality of Natural Language: A Neural Machine Translation Case Study

• Things not Written in Text: Exploring Spatial Commonsense from Visual Signals

• Toward Interpretable Semantic Textual Similarity via Optimal Transportbased Contrastive Sentence Learning

• Transformers in the loop: Polarity in neural models of language

• Upstream Mitigation Is Not All You Need: Testing the Bias Transfer Hypothesis in Pre-Trained Language Models

• When did you become so smart, oh wise one?! Sarcasm Explanation in Multi-modal Multi-party Dialogues

• Where to Go for the Holidays: Towards Mixed-Type Dialogs for Clarification of User Goals

• Which side are you on? Insider-Outsider classification in conspiracy theoretic social media

• Word Order Does Matter and Shuffled Language Models Know It

Language Model【语言模型】

模型结构

• ABC: Attention with Bounded-memory Control

• AdapLeR: Speeding up Inference by Adaptive Length Reduction

• AlephBERT: Language Model Pre-training and Evaluation from Sub-Word to Sentence Level

• Better Language Model with Hypernym Class Prediction

• CAMERO: Consistency Regularized Ensemble of Perturbed Language Models with Weight Sharing

• ClarET: Pre-training a Correlation-Aware Context-To-Event Transformer for Event-Centric Generation and Classification

• ClusterFormer: Neural Clustering Attention for Efficient and Effective Transformer

• Dependency-based Mixture Language Models

• E-LANG: Energy-Based Joint Inferencing of Super and Swift Language Models

• EPT-X: An Expression-Pointer Transformer model that generates eXplanations for numbers

• Exploring and Adapting Chinese GPT to Pinyin Input Method

• Few-Shot Tabular Data Enrichment Using Fine-Tuned Transformer Architectures

• Fine- and Coarse-Granularity Hybrid Self-Attention for Efficient BERT

• FORTAP: Using Formulas for Numerical-Reasoning-Aware Table Pretraining

• Fully Hyperbolic Neural Networks

• GLM: General Language Model Pretraining with Autoregressive Blank Infilling

• infty-former: Infinite Memory Transformer

• KinyaBERT: a Morphology-aware Kinyarwanda Language Model

• Knowledge Neurons in Pretrained Transformers

• LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding

• Long-range Sequence Modeling with Predictable Sparse Attention

• Low-Rank Softmax Can Have Unargmaxable Classes in Theory but Rarely in Practice

• Making Transformers Solve Compositional Tasks

• Pyramid-BERT: Reducing Complexity via Successive Core-set based Token Selection

• SkipBERT: Efficient Inference with Shallow Layer Skipping

• Sparsifying Transformer Models with Trainable Representation Pooling

• StableMoE: Stable Routing Strategy for Mixture of Experts

• TableFormer: Robust Transformer Modeling for Table-Text Encoding

• Transkimmer: Transformer Learns to Layer-wise Skim

训练策略

• The Trade-offs of Domain Adaptation for Neural Language Models

• A Simple Hash-Based Early Exiting Approach For Language Understanding and Generation

• Feeding What You Need by Understanding What You Learned

• Distinguishing Non-natural from Natural Adversarial Samples for More Robust Pre-trained Language Model

• Training Data is More Valuable than You Think: A Simple and Effective Method by Retrieving from Training Data

• ELLE: Efficient Lifelong Pre-training for Emerging Data

• LinkBERT: Pretraining Language Models with Document Links

• CoCoLM: Complex Commonsense Enhanced Language Model with Discourse Relations

• Coherence boosting: When your pretrained language model is not paying enough attention

• Feeding What You Need by Understanding What You Learned

• LinkBERT: Pretraining Language Models with Document Links

• MarkupLM: Pre-training of Text and Markup Language for Visually Rich Document Understanding

• Sparse Progressive Distillation: Resolving Overfitting under Pretrain-andFinetune Paradigm

• Token Dropping for Efficient BERT Pretraining

• XLM-E: Cross-lingual Language Model Pre-training via ELECTRA

模型压缩

• Compression of Generative Pre-trained Language Models via Quantization

• BERT Learns to Teach: Knowledge Distillation with Meta Learning

• Multi-Granularity Structural Knowledge Distillation for Language Model Compression

• Structured Pruning Learns Compact and Accurate Models

微调策略

• A Closer Look at How Fine-tuning Changes BERT

• A Good Prompt Is Worth Millions of Parameters: Low-resource Promptbased Learning for Vision-Language Models

• Adversarial Soft Prompt Tuning for Cross-Domain Sentiment Analysis

• An Information-theoretic Approach to Prompt Engineering Without Ground Truth Labels

• Are Prompt-based Models Clueless?

• bert2BERT: Towards Reusable Pretrained Language Models

• CogTaskonomy: Cognitively Inspired Task Taxonomy Is Beneficial to Transfer Learning in NLP

• Composable Sparse Fine-Tuning for Cross-Lingual Transfer

• ConTinTin: Continual Learning from Task Instructions

• Cross-Task Generalization via Natural Language Crowdsourcing Instructions

• Efficient Unsupervised Sentence Compression by Fine-tuning Transformers with Reinforcement Learning

• Enhancing Cross-lingual Natural Language Inference by Prompt-learning from Cross-lingual Templates

• Fantastically Ordered Prompts and Where to Find Them: Overcoming FewShot Prompt Order Sensitivity

• Few-Shot Learning with Siamese Networks and Label Tuning

• Knowledgeable Prompt-tuning: Incorporating Knowledge into Prompt Verbalizer for Text Classification

• On Continual Model Refinement in Out-of-Distribution Data Streams

• Overcoming Catastrophic Forgetting beyond Continual Learning: Balanced Training for Neural Machine Translation

• PPT: Pre-trained Prompt Tuning for Few-shot Learning

• Prompt-Based Rule Discovery and Boosting for Interactive WeaklySupervised Learning

• Prompt for Extraction? PAIE: Prompting Argument Interaction for Event Argument Extraction

• Prompt-free and Efficient Few-shot Learning with Language Models

• Prototypical Verbalizer for Prompt-based Few-shot Tuning

• Turning Tables: Generating Examples from Semi-structured Tables for Endowing Language Models with Reasoning Skills

• UniPELT: A Unified Framework for Parameter-Efficient Language Model Tuning