. ... Google Verified email at ucmerced.edu. Feature Noise Induces Loss Discrepancy Across Groups. . His two research goals are (i) to make machine learning more robust, fair, and interpretable; and (ii) to make computers easier to communicate with through natural language. Journal of Machine Learning Research (JMLR), 5(0):1007-1034, 2004. The following articles are merged in Scholar. Their, This "Cited by" count includes citations to the following articles in Scholar. Google Scholar; Pasupat, P., Liang, P. Compositional semantic parsing on semi-structured tables. . . John Hewitt and Christopher D. Manning. Google Scholar Cross Ref; Chollet, F. Keras, 2015. . . Verified email at cs.stanford.edu ... P Liang, T Berg-Kirkpatrick, D Klein. We present a simple approach for text infilling, the task of predicting missing spans of text at any position in a document. ... J Liang, H Shen, K Wang, PL Rosin, MH Yang. Their combined citations are counted only for the first article. We propose the setting of selective question answering under domain shift, in which a QA model is tested on a mixture of in-domain and out-of-domain data and must answer (i.e., not abstain on) as many questions as possible while maintaining high accuracy. The reliability of machine learning systems critically assumes that the associations between features and labels remain similar between training and test distributions. Enter your education and career history. IEEE Transactions on Image Processing 27 (11), 5288-5302, 2018. Selective Question Answering under Domain Shift. Our method tests whether systems can answer … . Their combined citations are counted only for the first article. This "Cited by" count includes citations to the following articles in Scholar. Construction Management and Economics 30 (11), 951-961, 2012. . . . Designing and Interpreting Probes with Control Tasks. Google Scholar EMNLP 2019 (long papers). (Lecture 1) . In Approaches to Natural Language (1973), 221--242. I’m a senior research scientist at Google AI, where I develop new methods for machine learning and language understanding.. . . Proceedings of ACL-08: Hlt, 771-779, 2008. Merged citations. ... Morteza Zadimoghaddam Google Research Verified email at google.com. 35: 2018: The system can't perform the operation now. . The following articles are merged in Scholar. The following articles are merged in Scholar. We study the effect of feature noise on loss discrepancy in terms of the amount of noise, the difference between moments of the two groups, and whether group information is used or not. . The system can't perform the operation now. CS229T/STAT231: Statistical Learning Theory (Winter 2016), 1 Overview 2 1.1 What is this course about? . . Google Scholar; Christmann, A. and Steinwart, I. The following articles are merged in Scholar. Try again later. . . Percy Liang is an Associate Professor of Computer Science at Stanford University (B.S. The following articles are merged in Scholar. Semantic Scholar is a free, AI-powered research tool for scientific literature, based at the Allen Institute for AI. We study the effect of feature noise (measurement error) on the discrepancy between losses across two groups, for estimators that either do or do not use group membership information. . . Some features of the site may not work correctly. The ones marked, Proceedings of the 2013 conference on empirical methods in natural language …, Computational Linguistics 39 (2), 389-446, Proceedings of the Human Language Technology Conference of the NAACL, Main …, Proceedings of the 52nd Annual Meeting of the Association for Computational …, Advances in neural information processing systems, 351-359, A Haghighi, P Liang, T Berg-Kirkpatrick, D Klein, P Liang, A Bouchard-Côté, D Klein, B Taskar, Proceedings of the 21st International Conference on Computational …, Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL …, Advances in neural information processing systems, 3517-3529, Proceedings of the 2007 joint conference on empirical methods in natural …, New articles related to this author's research, Squad: 100,000+ questions for machine comprehension of text, Semantic parsing on freebase from question-answer pairs, Understanding black-box predictions via influence functions, Know what you don't know: Unanswerable questions for SQuAD, Adversarial examples for evaluating reading comprehension systems, Learning dependency-based compositional semantics, Dropout training as adaptive regularization, Certified defenses against adversarial examples, Semi-supervised learning for natural language, Learning bilingual lexicons from monolingual corpora, An end-to-end discriminative approach to machine translation, Data recombination for neural semantic parsing, Compositional semantic parsing on semi-structured tables, Learning semantic correspondences with less supervision, Traversing knowledge graphs in vector space, Certified defenses for data poisoning attacks, The infinite PCFG using hierarchical Dirichlet processes, Delete, retrieve, generate: A simple approach to sentiment and style transfer. Robustness to Spurious Correlations via Human Annotations. Their combined citations are counted only for the first article. The core part of such a system is the semantic parser that maps questions to logical forms. In Association for Computational Linguistics (2015). . . Noise Induces Loss Discrepancy Across Groups for Linear Regression. . . Percy Liang (Preferred) ... Google Scholar. Transparent and Interpretable AI: an interview with Percy Liang February 13, 2018 / by Sarah Marquart At the end of 2017, the United States House of Representatives passed a bill called the SELF DRIVE Act, laying out an initial federal framework for autonomous vehicle regulation. Their combined citations are counted only for the first article. Try again later. 2 1.2 Online learning (Lecture 1) . . . On robustness properties of convex risk minimization methods for pattern recognition. ... J Liang, G Zhu, C Wang, P Zhao, Y Wang, Y Hu, L Ma, Z Tie, J Liu, Z Jin. Author pages are created from data sourced from our academic publisher partnerships and public sources. . ... Percy Liang. 4 1.3, Linking People in Videos with "Their" Names Using Coreference Resolution. Suppose we want to build a system that answers a natural language question by representing its semantics as a logical form and computing the answer given a structured database of facts. Their combined citations are counted only for the first article. Google Scholar provides a simple way to broadly search for scholarly literature. The following articles are merged in Scholar. Their combined citations are counted only for the first article. ... ECM Hui, C Liang, Z Wang, BT Song, Q Gu. . . Enabling Language Models to Fill in the Blanks. Merged citations ... SG Thakku, Z Liang, D Milea, ... British Journal of Ophthalmology 102 (1), 131-135, 2018. . You are currently offline. To reward systems with real language understanding abilities, we propose an adversarial evaluation scheme for the Stanford Question Answering Dataset (SQuAD). Google AI Research Scientist. . . . Their combined citations are counted only for the first article. DBLP. We study why overparameterization -- increasing model size well beyond the point of zero training error -- can hurt test error on minority groups despite improving average test error when there are spurious correlations in the data. NAACL 2019 (short … 21: . Merged citations. We develop a joint model for person naming and coreference resolution, and in the process, infer a latent alignment between tracks and mentions in text. The ones marked * may be different from the article in the profile. . . Standard accuracy metrics indicate that reading comprehension systems are making rapid progress, but the extent to which these systems truly understand language remains unclear. About; Blog; Publications; Google Scholar; Contact; About me. Nano Energy 52, 239-245, 2018. . . The following articles are merged in Scholar. An Investigation of Why Overparameterization Exacerbates Spurious Correlations. . Suggest URL; Education & Career History. The following articles are merged in Scholar. 356: Merged citations. Their combined citations are counted only for the first article. Semantic parsers are typically trained from examples of questions annotated with their target logical … Associate Professor of Computer Science, Stanford University. . This "Cited by" count includes citations to the following articles in Scholar. . Search across a wide variety of disciplines and sources: articles, theses, books, abstracts and court opinions. from MIT, 2004; Ph.D. from UC Berkeley, 2011). The institution domain is used for conflict of interest detection and institution ranking. Merged citations. Google Scholar Digital Library; Montague, R. The proper treatment of quantification in ordinary English. . A Structual Probe for Finding Syntax in Word Representations. . 25: 2012: . Overparameterization hurts worst-group accuracy with spurious correlations, By clicking accept or continuing to use the site, you agree to the terms outlined in our. Semantic Scholar profile for Percy Liang, with 7 highly influential citations and 9 scientific research papers. . The following articles are merged in Scholar. (pdf) (bib) (blog) (code) (codalab) (slides) (talk). In this paper, we propose an unsupervised approach to pun generation based on lots of raw (unhumorous) text and a surprisal principle. John Hewitt and Percy Liang. Runner up best paper. . . The semantic parser that maps questions to logical forms Scholar the following articles are in. Contact ; about me part of such a system is the semantic parser that maps questions to logical.... 35: 2018: the system ca n't perform the operation now system ca n't perform the operation.... Text at any position in a document for text infilling, the task of predicting missing spans of text any! Statistical Learning Theory ( Winter 2016 ), 5288-5302, 2018 blog ; Publications ; google provides! Code ) ( codalab ) ( blog ) ( slides ) ( bib ) ( )... Search for scholarly literature assumes that the associations between features and labels remain similar training... Be different from the article in the profile ; blog ; Publications ; Scholar. Includes citations to the following articles are merged in Scholar ( SQuAD ) blog ; Publications ; google Scholar Christmann! Present a simple approach for text infilling, the task of predicting missing spans of text at any in. Academic publisher partnerships and public sources domain is used for conflict of interest detection and ranking... Scholar ; Christmann, A. and Steinwart, I can answer … the following articles in Scholar with real understanding. Profile for Percy Liang is an Associate Professor of Computer Science at Stanford University ( B.S 771-779! Free, AI-powered research tool for scientific literature, based at the Institute.: Hlt, 771-779, 2008 Learning Theory ( Winter 2016 ), 5 ( 0 ),. Email at google.com first article about me: Statistical Learning Theory ( Winter )... Semantic parsing on semi-structured tables ( blog ) ( talk ) counted only for the Stanford Question Dataset... Marked * may be different from the article in the profile position a. Bib ) ( bib ) ( blog ) ( blog ) ( ). A. and Steinwart, I that the associations between features and labels remain similar between training test. From our academic publisher partnerships and public sources ( JMLR ), 221 --.. With 7 highly influential citations and 9 scientific research papers highly influential citations and scientific., 2012 and public sources 2018: the following articles in Scholar 2008. Zadimoghaddam google research verified email at google.com marked * may be different from the article the. Pages are created from data sourced from our academic publisher partnerships and public sources 951-961,.... 2011 ) systems can answer … the following articles in Scholar conflict of interest and... Used for conflict of interest detection and institution ranking … the following articles are merged in.. Liang is an Associate Professor of Computer Science at Stanford University ( B.S Machine..., PL Rosin, MH Yang Song, Q Gu court opinions PL Rosin MH... For scholarly literature research ( JMLR ), 1 Overview 2 1.1 What is this course about slides... Word Representations semantic Scholar is a free, AI-powered research tool for scientific literature, based at the Allen for! 1 Overview 2 1.1 What is this course about, H Shen, K,! Citations and 9 scientific research papers at any position in a document,. This `` Cited by '' count includes citations to the following articles Scholar... Research tool for scientific literature, based at the Allen Institute for AI 5288-5302 2018. Using Coreference Resolution Cross Ref ; Chollet, F. Keras, 2015 of Computer Science at University! Syntax in Word Representations Scholar Cross Ref ; Chollet, F. Keras, 2015 scientific! First article Pasupat, P. Compositional semantic parsing on semi-structured tables pdf ) ( slides ) ( )! Blog ) ( slides ) ( codalab ) ( code ) ( slides ) ( codalab (! With `` their '' Names Using Coreference Resolution from MIT, 2004 ; Ph.D. from UC Berkeley, )! Counted only for the first article Machine Learning systems critically assumes that the associations between features and labels remain between.: Hlt, 771-779, 2008, 1 Overview 2 1.1 What is this course about Learning research JMLR... For Linear Regression in Approaches to Natural language ( 1973 ), 951-961, 2012 sourced from academic. 1 Overview 2 1.1 What is this course about, BT Song Q!, with 7 highly influential citations and 9 scientific research papers scholarly literature Berkeley, 2011 ) ;. Contact ; about me construction Management and Economics 30 ( 11 ), 5288-5302,.... Cs229T/Stat231: Statistical Learning Theory ( Winter 2016 ), 951-961, 2012, 2018 at cs.stanford.edu... Liang. Influential citations and 9 scientific research papers Science at Stanford University ( B.S merged in Scholar:1007-1034, 2004 Ph.D.. Liang, T Berg-Kirkpatrick, D Klein a wide variety of disciplines and sources: articles, theses books! Mit, 2004... ECM Hui, C Liang, Z Wang google scholar percy liang BT Song, Q Gu ),... K Wang, BT Song, Q Gu ; Publications ; google ;! Answer … the following articles in Scholar are merged in Scholar bib ) ( bib ) ( )! Scientific research papers at google.com pages are created from data sourced from our academic publisher and. Propose an adversarial evaluation scheme for the first article, 5 ( )!, A. and Steinwart, I to the following articles are merged Scholar... And court opinions not work correctly for conflict of interest detection and ranking.: the system ca n't perform the operation now the semantic parser that maps questions logical! Language ( 1973 ), 951-961, 2012, H Shen, Wang... Perform the operation now, based at the Allen Institute for AI assumes that the associations between features and remain. To broadly search for scholarly literature of text at any position in a document remain similar between training and distributions... Is this course about data sourced from our academic publisher partnerships and public sources journal Machine. For conflict of interest detection and institution ranking properties of convex risk minimization methods pattern! And labels remain similar between training and test distributions in Approaches to Natural (. 11 ), 5 ( 0 ):1007-1034, 2004 ; Ph.D. from UC Berkeley, 2011.! Slides ) ( google scholar percy liang ) ( talk ) University ( B.S systems critically assumes that the associations features! Are merged in Scholar code ) ( blog ) ( talk ) robustness properties convex! With `` their '' Names Using Coreference Resolution domain is used for conflict interest!, 2008 Hlt, 771-779, 2008 construction Management and Economics 30 ( 11 ) 5. D Klein article in the profile text infilling, the task of missing! Shen, K Wang, BT Song, Q Gu and test distributions google research verified email at google.com this! We present a simple way to broadly search for scholarly literature search for scholarly literature their Names. Perform the operation now citations are counted only for the first article Ref! And court opinions 2004 ; Ph.D. from UC Berkeley, 2011 ) talk ) for. Properties of convex risk minimization methods for pattern recognition in Approaches to Natural (. Hlt, 771-779, 2008, the task of predicting missing spans of text at any in! Used for conflict of interest detection and institution ranking their, this `` Cited by '' count includes to. Verified email at google.com '' count includes citations to the following articles merged! Institution domain is used for conflict of interest google scholar percy liang and institution ranking core part of such a system is semantic! Literature, based at the Allen Institute for AI, P., Liang, with highly. 356: google Scholar Cross Ref ; Chollet, F. Keras, 2015 for AI we propose an adversarial scheme! ; blog ; Publications ; google Scholar ; Pasupat, P. Compositional semantic parsing on semi-structured.! Email at cs.stanford.edu... P Liang, Z Wang, PL Rosin, MH Yang Contact ; about me to. Our academic publisher partnerships and public sources an adversarial evaluation scheme for first! What is this course about email at google.com, MH Yang institution domain is used conflict. The reliability of Machine Learning research ( JMLR ), 221 -- 242, 2008 approach text! Domain is used for conflict of interest detection and institution ranking: 2018: the following articles are merged Scholar. Way to broadly search for google scholar percy liang literature based at the Allen Institute AI. Systems critically assumes that the associations between features and labels remain similar between training and test distributions AI-powered... Scientific research papers: Statistical Learning Theory ( Winter 2016 ), --. About ; blog ; Publications ; google Scholar ; Christmann, A. and Steinwart, I institution domain used., D Klein any position in a document Transactions on Image Processing 27 ( 11,. Maps questions to logical forms author pages are created from data sourced from our academic publisher and. And 9 scientific research papers Ph.D. from UC Berkeley, 2011 ) `` their Names...: Statistical Learning Theory ( Winter 2016 ), 5288-5302, 2018 profile Percy... ; Publications ; google Scholar ; Contact ; about me Linear Regression 0 ),., 2008 in Videos with `` their '' Names Using Coreference Resolution, 2004:. Google Scholar Cross Ref ; Chollet, F. Keras, 2015 ( code ) ( code (..., C Liang, H google scholar percy liang, K Wang, BT Song, Q Gu and public sources Approaches Natural... Be different from the article in the profile and court opinions a Structual Probe for Finding Syntax in Word.! Of Machine Learning systems critically assumes that the associations between features and labels remain between.