![自然语言处理学术速递[2021.10.29]](data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'></svg>)
自然语言处理学术速递[2021.10.29]
访问arxivdaily.com获取含摘要速递，更有收藏、搜索等功能，涵盖CS|物理|数学|经济|统计|金融|生物|电气领域
同步公众号:arXiv每日学术速递，欢迎关注
cs.CL 方向，今日共计34篇
Transformer(2篇)
【1】 Pruning Attention Heads of Transformer Models Using A* Search: A Novel Approach to Compress Big NLP Architectures
基于A*搜索的Transformer模型关注头剪枝：一种压缩大型NLP体系结构的新方法
链接：https://arxiv.org/abs/2110.15225
作者：Archit Parnami,Rahul Singh,Tarun Joshi
机构：University of North Carolina at Charlotte,Corporate Model Risk, Wells Fargo, USA
备注：23 Pages, 18 figures, 3 tables
【2】 Detecting Dementia from Speech and Transcripts using Transformers
使用Transformer从语音和抄本中检测痴呆症
链接：https://arxiv.org/abs/2110.14769
作者：Loukas Ilias,Dimitris Askounis,John Psarras
机构：School of Electrical and Computer Engineering, National Technical University of Athens, Zografou, Athens
备注：This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
BERT(1篇)
【1】 When is BERT Multilingual? Isolating Crucial Ingredients for Cross-lingual Transfer
BERT什么时候会说多种语言？分离跨语言迁移的关键成分
链接：https://arxiv.org/abs/2110.14782
作者：Ameet Deshpande,Partha Talukdar,Karthik Narasimhan
机构：Department of Computer Science, Princeton University, USA, Google Research, India
QA|VQA|问答|对话(1篇)
【1】 Multi-stage Clarification in Conversational AI: The case of Question-Answering Dialogue Systems
会话人工智能中的多阶段澄清：问答对话系统的案例
链接：https://arxiv.org/abs/2110.15235
作者：Hadrien Lautraite,Nada Naji,Louis Marceau,Marc Queudot,Eric Charton
机构：National Bank Of Canada, de la Gauchetière, Montréal, Québec
备注：9 pages, 3 figures, 2 tables
机器翻译(1篇)
【1】 Fine Grained Human Evaluation for English-to-Chinese Machine Translation: A Case Study on Scientific Text
面向英汉机器翻译的细粒度人类评价--以科技文本为例
链接：https://arxiv.org/abs/2110.14766
作者：Ming Liu,He Zhang,Guanhao Wu
机构： Deakin University, Melbourne, Australia, Zhongtukexin Co. Ltd. , Beijing, China
备注：12 pages, 3 tables
语义分析(1篇)
【1】 SenTag: a Web-based Tool for Semantic Annotation of Textual Documents
SenTag：一种基于Web的文本语义标注工具
链接：https://arxiv.org/abs/2110.15062
作者：Andrea Loreggia,Simone Mosco,Alberto Zerbinati
机构： European University Institute, University of Padova
推理|分析|理解|解释(3篇)
【1】 An Analysis of Programming Course Evaluations Before and After the Introduction of an Autograder
程序设计课程在引进自动评分机前后的评价分析
链接：https://arxiv.org/abs/2110.15134
作者：Gerhard Hagerer,Laura Lahesoo,Miriam Anschütz,Stephan Krusche,Georg Groh
机构：Department of Informatics, Technical University of Munich, Boltzmannstraße , Garching nearby Munich, Germany
备注：None
【2】 Towards Fine-Grained Reasoning for Fake News Detection
面向假新闻识别的细粒度推理研究
链接：https://arxiv.org/abs/2110.15064
作者：Yiqiao Jin,Xiting Wang,Ruichao Yang,Yizhou Sun,Wei Wang,Hao Liao,Xing Xie
机构： University of California, Los Angeles, Microsoft Research Asia, Hong Kong Baptist University, Shenzhen University
备注：9 pages, 5 figures
【3】 Empirical Analysis of Korean Public AI Hub Parallel Corpora and in-depth Analysis using LIWC
韩国公共人工智能枢纽平行语料库实证分析及LIWC深度分析
链接：https://arxiv.org/abs/2110.15023
作者：Chanjun Park,Midan Shim,Sugyeong Eo,Seolhwa Lee,Jaehyung Seo,Hyeonseok Moon,Heuiseok Lim
机构：Korea University,Kyunghee University
GAN|对抗|攻击|生成相关(3篇)
【1】 Bridge the Gap Between CV and NLP! A Gradient-based Textual Adversarial Attack Framework
弥合简历和NLP之间的鸿沟！一种基于梯度的文本对抗性攻击框架
链接：https://arxiv.org/abs/2110.15317
作者：Lifan Yuan,Yichi Zhang,Yangyi Chen,Wei Wei
机构：Cognitive Computing and Intelligent Information Processing Laboratory, School of Computer Science and Technology Huazhong University of Science and Tehchnology
备注：Work on progress
【2】 Generating Table Vector Representations
生成表向量表示
链接：https://arxiv.org/abs/2110.15132
作者：Aneta Koleva,Martin Ringsquandl,Mitchell Joblin,Volker Tresp
机构：Siemens, Otto-Hahn-Ring , Munich, Germany, Ludwig Maximilian University of Munich, Geschwister-Scholl-Platz , Munich, Germany
备注：Accepted at DL4KF@ISWC
【3】 Preventing posterior collapse in variational autoencoders for text generation via decoder regularization
通过解码器正则化防止用于文本生成的可变自动编码器的后部崩溃
链接：https://arxiv.org/abs/2110.14945
作者：Alban Petit,Caio Corro
机构：Université Paris-Saclay, CNRS, LISN, Orsay, France
备注：Accepted at NeurIPS 2021 Workshop DGMs Applications
检测相关(3篇)
【1】 Confounds and Overestimations in Fake Review Detection: Experimentally Controlling for Product-Ownership and Data-Origin
虚假评论检测中的混淆和高估：对产品所有权和数据来源的实验控制
链接：https://arxiv.org/abs/2110.15130
作者：Felix Soldner,Bennett Kleinberg,Shane Johnson
机构：GESIS – Leibniz Institute for the Social Science, Köln, Germany, Department of Security and Crime Science & Dawes Centre for Future Crime, University College London, UK, Department of Methodology and Statistics, Tilburg University, The Netherlands
备注：This paper is a pre-print
【2】 Combining Vagueness Detection with Deep Learning to Identify Fake News
模糊检测与深度学习相结合的假新闻识别
链接：https://arxiv.org/abs/2110.14780
作者：Paul Guélorget,Benjamin Icard,Guillaume Gadek,Souhir Ghabiche,Sylvain Gatepaille,Ghislain Atemezing,Paul Égré
机构：∗Advanced Information Processing, Airbus, France, †Institut Jean-Nicod, CNRS, ENS, EHESS, PSL University, France, ‡Mondeca, Paris, France
备注：Paper to appear in the Proceedings of the 24th International Conference on Information Fusion. Johannesburg
【3】 Anomaly-Injected Deep Support Vector Data Description for Text Outlier Detection
用于文本孤立点检测的异常注入深度支持向量数据描述
链接：https://arxiv.org/abs/2110.14729
作者：Zeyu You,Yichu Zhou,Tao Yang,Wei Fan
机构：Afiliations ,Association for Tencent America, Park Blvd, Palo Alto, California , Association for University of Utah, Presidents’ Cir, Salt Lake City, UT
备注：11 pages, 5 figures, 3 tables
识别/分类(4篇)
【1】 TEXTOIR: An Integrated and Visualized Platform for Text Open Intent Recognition
TEXTOIR：一个集成的可视化文本开放意图识别平台
链接：https://arxiv.org/abs/2110.15063
作者：Hanlei Zhang,Xiaoteng Li,Hua Xu,Panpan Zhang,Kang Zhao,Kai Gao
机构：State Key Laboratory of Intelligent Technology and Systems, Department of Computer Science and Technology, Tsinghua University, School of Information Science and Engineering, Hebei University of Science and Technology
备注：Published in ACL 2021, demo paper
【2】 End-to-End Speech Emotion Recognition: Challenges of Real-Life Emergency Call Centers Data Recordings
端到端语音情感识别：现实生活中紧急呼叫中心数据记录的挑战
链接：https://arxiv.org/abs/2110.14957
作者：Théo Deschamps-Berger,Lori Lamel,Laurence Devillers
机构：LISN, CNRS, Orsay, France
备注：None
【3】 Hate Speech Classifiers Learn Human-Like Social Stereotypes
仇恨言语分类器学习与人类相似的社会刻板印象
链接：https://arxiv.org/abs/2110.14839
作者：Aida Mostafazadeh Davani,Mohammad Atari,Brendan Kennedy,Morteza Dehghani
机构：Department of Computer Science, University of Southern California, Department of Psychology, University of Southern California, Brain and Creativity Institute, University of Southern California, Author Note
【4】 Generalized Funnelling: Ensemble Learning and Heterogeneous Document Embeddings for Cross-Lingual Text Classification
广义漏斗：用于跨语言文本分类的集成学习和异构文档嵌入
链接：https://arxiv.org/abs/2110.14764
作者：Alejandro Moreo,Andrea Pedrotti,Fabrizio Sebastiani
机构：Ricerche, Pisa, Italy, Funnelling (Fun) is a recently proposed method for cross-lingual text classification (CLTC) based on a two-tier, learning ensemble for heterogeneous transfer learning (HTL). In this ensemble method,st-tier classifiers
Zero/Few/One-Shot|迁移|自适应(1篇)
【1】 Adaptive Multimodal and Multisensory Empathic Technologies for Enhanced Human Communication
增强人类交流的自适应多模态和多感官共情技术
链接：https://arxiv.org/abs/2110.15054
作者：Roxana Girju
机构：Department of Linguistics, Department of Computer Science, Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana IL
备注：None
Word2Vec|文本|单词(1篇)
【1】 Word-level confidence estimation for RNN transducers
RNN换能器的词级置信度估计
链接：https://arxiv.org/abs/2110.15222
作者：Mingqiu Wang,Hagen Soltau,Laurent El Shafey,Izhak Shafran
机构：Google Research
备注：None
其他神经网络|深度学习|模型|建模(8篇)
【1】 Learning to Ground Multi-Agent Communication with Autoencoders
使用自动编码器学习多Agent通信的接地
链接：https://arxiv.org/abs/2110.15349
作者：Toru Lin,Minyoung Huh,Chris Stauffer,Ser-Nam Lim,Phillip Isola
机构：MIT CSAIL, Facebook AI
备注：Project page, code, and videos can be found at this https URL
【2】 Cognitive network science quantifies feelings expressed in suicide letters and Reddit mental health communities
认知网络科学将自杀信件和Reddit心理健康社区中表达的情感量化
链接：https://arxiv.org/abs/2110.15269
作者：Simmi Marina Joseph,Salvatore Citraro,Virginia Morini,Giulio Rossetti,Massimo Stella
机构：CogNosco Lab, Department of Computer Science, University of Exeter, Exeter, UK, Department of Computer Science, University of Pisa, Largo Bruno Pontecorvo, Pisa, KDD-Lab (ISTI-CNR), G. Moruzzi, Pisa
【3】 Abstract, Rationale, Stance: A Joint Model for Scientific Claim Verification
抽象、理据、立场：科学索赔验证的联合模型
链接：https://arxiv.org/abs/2110.15116
作者：Zhiwei Zhang,Jiyi Li,Fumiyo Fukumoto,Yanming Ye
机构：Hangzhou Dianzi University, Hangzhou, China, University of Yamanashi, Kofu, Japan
备注：EMNLP 2021
【4】 Semi-Siamese Bi-encoder Neural Ranking Model Using Lightweight Fine-Tuning
基于轻量级微调的半暹罗双编码器神经网络排序模型
链接：https://arxiv.org/abs/2110.14943
作者：Euna Jung,Jaekeol Choi,Wonjong Rhee
机构：Seoul National University, Seoul, Korea, & Naver Corp., GSCST, GSAI, AIIS
备注：10 pages, 4 figures
【5】 Colossal-AI: A Unified Deep Learning System For Large-Scale Parallel Training
面向大规模并行训练的统一深度学习系统Lossal-AI
链接：https://arxiv.org/abs/2110.14883
作者：Zhengda Bian,Hongxin Liu,Boxiang Wang,Haichen Huang,Yongbin Li,Chuanrui Wang,Fan Cui,Yang You
机构：HPC-AI Technology Inc., National University of Singapore (NUS)
【6】 A Sequence to Sequence Model for Extracting Multiple Product Name Entities from Dialog
用于从对话框中提取多个产品名称实体的序列到序列模型
链接：https://arxiv.org/abs/2110.14843
作者：Praneeth Gubbala,Xuan Zhang
机构：Walmart Global Tech
备注：WeCNLP 2021 camera-ready
【7】 Towards Realistic Single-Task Continuous Learning Research for NER
面向NER的面向现实的单任务持续学习研究
链接：https://arxiv.org/abs/2110.14694
作者：Justin Payan,Yuval Merhav,He Xie,Satyapriya Krishna,Anil Ramakrishna,Mukund Sridhar,Rahul Gupta
机构：University of Massachusetts Amherst, MA, USA, Amazon Alexa AI, MA, USA
备注：11 pages, 2 figures, Findings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) (short paper), November 2021
【8】 Emoji-aware Co-attention Network with EmoGraph2vec Model for Sentiment Anaylsis
基于EmoGraph2vec模型的表情感知共同关注网络情感分析
链接：https://arxiv.org/abs/2110.14636
作者：Xiaowei Yuan,Jingyuan Hu,Xiaodan Zhang,Honglei Lv,Hao Liu
机构：Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China, School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China
备注：arXiv admin note: text overlap with arXiv:2110.14227
其他(5篇)
【1】 ÚFAL at MultiLexNorm 2021: Improving Multilingual Lexical Normalization by Fine-tuning ByT5
在MultiLexNorm2021上使用FAL：通过T5的微调改进多语言词汇规范化
链接：https://arxiv.org/abs/2110.15248
作者：David Samuel,Milan Straka
机构：Charles University, Institute of Formal and Applied Linguistics
备注：Accepted to W-NUT 2021
【2】 An Add-On for Empowering Google Forms to be an Automatic Question Generator in Online Assessments
支持Google Forms成为在线评估中的自动问题生成器的附加组件
链接：https://arxiv.org/abs/2110.15220
作者：Pornpat Sirithumgul,Pimpaka Prasertsilp,Lorne Olfman
机构：Department of Computer Engineering, Rajamangala University of Technology Phra Nakhon, Thailand., School of Science and Technology, Sukhothai Thammathirat Open University, Thailand.
【3】 BERTian Poetics: Constrained Composition with Masked LMs
贝尔田诗学：蒙面LMS的限制性写作
链接：https://arxiv.org/abs/2110.15181
作者：Christopher Akiki,Martin Potthast
备注：Accepted as a poster at the 2021 NeurIPS Workshop on Machine Learning for Creativity and Design
【4】 Diversity-Driven Combination for Grammatical Error Correction
多样性驱动的组合语法纠错方法
链接：https://arxiv.org/abs/2110.15149
作者：Wenjuan Han,Hwee Tou Ng
机构：Department of Computer Science, National University of Singapore
备注：Accepted by ICTAI 2021
【5】 Dynamic Review-based Recommenders
基于动态评论的推荐器
链接：https://arxiv.org/abs/2110.14747
作者：Kostadin Cvejoski,Ramses J. Sanchez,Christian Bauckhage,Cesar Ojeda
机构：∗Competence Center Machine Learning Rhine-Ruhr, †B-IT, University of Bonn, Bonn, Germany, ‡Fraunhofer Center for Machine Learning and Fraunhofer IAIS, Germany, §Berlin Center for Machine Learning and TU Berlin, Berlin, Germany
备注：6pages, Published at International Data Science Conference 2021 (iDSC21)
机器翻译，仅供参考
访问arxivdaily.com获取含摘要速递，更有收藏、搜索等功能，涵盖CS|物理|数学|经济|统计|金融|生物|电气领域
同步公众号:arXiv每日学术速递，欢迎关注
