Publications
2017 |
Xu, Hainan; Koehn, Philipp Zipporah: a Fast and Scalable Data Cleaning System for Noisy Web-Crawled Parallel Corpora Inproceedings Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pp. 2935–2940, Association for Computational Linguistics, Copenhagen, Denmark, 2017. @inproceedings{xu-koehn:2017:EMNLP2017, title = {Zipporah: a Fast and Scalable Data Cleaning System for Noisy Web-Crawled Parallel Corpora}, author = {Hainan Xu and Philipp Koehn}, url = {https://www.aclweb.org/anthology/D17-1318}, year = {2017}, date = {2017-09-01}, booktitle = {Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing}, pages = {2935--2940}, publisher = {Association for Computational Linguistics}, address = {Copenhagen, Denmark}, abstract = {We introduce Zipporah, a fast and scalable data cleaning system. We propose a novel type of bag-of-words translation feature, and train logistic regression models to classify good data and synthetic noisy data in the proposed feature space. The trained model is used to score parallel sentences in the data pool for selection. As shown in experiments, Zipporah selects a high-quality parallel corpus from a large, mixed quality data pool. In particular, for one noisy dataset, Zipporah achieves a 2.1 BLEU score improvement with using 1/5 of the data over using the entire corpus.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We introduce Zipporah, a fast and scalable data cleaning system. We propose a novel type of bag-of-words translation feature, and train logistic regression models to classify good data and synthetic noisy data in the proposed feature space. The trained model is used to score parallel sentences in the data pool for selection. As shown in experiments, Zipporah selects a high-quality parallel corpus from a large, mixed quality data pool. In particular, for one noisy dataset, Zipporah achieves a 2.1 BLEU score improvement with using 1/5 of the data over using the entire corpus. |
Bojar, Ondřej; Chatterjee, Rajen; Federmann, Christian; Graham, Yvette; Haddow, Barry; Huang, Shujian; Huck, Matthias; Koehn, Philipp; Liu, Qun; Logacheva, Varvara; Monz, Christof; Negri, Matteo; Post, Matt; Rubino, Raphael; Specia, Lucia; Turchi, Marco Findings of the 2017 Conference on Machine Translation (WMT17) Inproceedings Proceedings of the Second Conference on Machine Translation, pp. 169–214, Association for Computational Linguistics, Copenhagen, Denmark, 2017. @inproceedings{bojar-EtAl:2017:WMT1, title = {Findings of the 2017 Conference on Machine Translation (WMT17)}, author = {Ondřej Bojar and Rajen Chatterjee and Christian Federmann and Yvette Graham and Barry Haddow and Shujian Huang and Matthias Huck and Philipp Koehn and Qun Liu and Varvara Logacheva and Christof Monz and Matteo Negri and Matt Post and Raphael Rubino and Lucia Specia and Marco Turchi}, url = {http://www.aclweb.org/anthology/W17-4717}, year = {2017}, date = {2017-09-01}, booktitle = {Proceedings of the Second Conference on Machine Translation}, pages = {169--214}, publisher = {Association for Computational Linguistics}, address = {Copenhagen, Denmark}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Nadejde, Maria; Reddy, Siva; Sennrich, Rico; Dwojak, Tomasz; Junczys-Dowmunt, Marcin; Koehn, Philipp; Birch, Alexandra Predicting Target Language CCG Supertags Improves Neural Machine Translation Inproceedings Proceedings of the Second Conference on Machine Translation, pp. 68–79, Association for Computational Linguistics, Copenhagen, Denmark, 2017. @inproceedings{nadejde-EtAl:2017:WMT, title = {Predicting Target Language CCG Supertags Improves Neural Machine Translation}, author = {Maria Nadejde and Siva Reddy and Rico Sennrich and Tomasz Dwojak and Marcin Junczys-Dowmunt and Philipp Koehn and Alexandra Birch}, url = {http://www.aclweb.org/anthology/W17-4707}, year = {2017}, date = {2017-09-01}, booktitle = {Proceedings of the Second Conference on Machine Translation}, pages = {68--79}, publisher = {Association for Computational Linguistics}, address = {Copenhagen, Denmark}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Ding, Shuoyang; Khayrallah, Huda; Koehn, Philipp; Post, Matt; Kumar, Gaurav; Duh, Kevin The JHU Machine Translation Systems for WMT 2017 Inproceedings Proceedings of the Second Conference on Machine Translation, pp. 276–282, Association for Computational Linguistics, Copenhagen, Denmark, 2017. @inproceedings{ding-EtAl:2017:WMT, title = {The JHU Machine Translation Systems for WMT 2017}, author = {Shuoyang Ding and Huda Khayrallah and Philipp Koehn and Matt Post and Gaurav Kumar and Kevin Duh}, url = {http://www.aclweb.org/anthology/W17-4724}, year = {2017}, date = {2017-09-01}, booktitle = {Proceedings of the Second Conference on Machine Translation}, pages = {276--282}, publisher = {Association for Computational Linguistics}, address = {Copenhagen, Denmark}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Koehn, Philipp; Knowles, Rebecca Six Challenges for Neural Machine Translation Inproceedings Proceedings of the First Workshop on Neural Machine Translation, pp. 28–39, Association for Computational Linguistics, Vancouver, 2017. @inproceedings{koehn-knowles:2017:NMT, title = {Six Challenges for Neural Machine Translation}, author = {Philipp Koehn and Rebecca Knowles}, url = {http://www.aclweb.org/anthology/W17-3204}, year = {2017}, date = {2017-08-01}, booktitle = {Proceedings of the First Workshop on Neural Machine Translation}, pages = {28--39}, publisher = {Association for Computational Linguistics}, address = {Vancouver}, abstract = {We explore six challenges for neural machine translation: domain mismatch, amount of training data, rare words, long sentences, word alignment, and beam search. We show both deficiencies and improvements over the quality of phrase-based statistical machine translation.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We explore six challenges for neural machine translation: domain mismatch, amount of training data, rare words, long sentences, word alignment, and beam search. We show both deficiencies and improvements over the quality of phrase-based statistical machine translation. |
Renduchintala, Adithya; Koehn, Philipp; Eisner, Jason Knowledge Tracing in Sequential Learning of Inflected Vocabulary Inproceedings Proceedings of the 21st Conference on Computational Natural Language Learning (CoNLL 2017), pp. 238–247, Association for Computational Linguistics, Vancouver, Canada, 2017. @inproceedings{renduchintala-koehn-eisner:2017:CoNLL, title = {Knowledge Tracing in Sequential Learning of Inflected Vocabulary}, author = {Adithya Renduchintala and Philipp Koehn and Jason Eisner}, url = {http://aclweb.org/anthology/K17-1025}, year = {2017}, date = {2017-08-01}, booktitle = {Proceedings of the 21st Conference on Computational Natural Language Learning (CoNLL 2017)}, pages = {238--247}, publisher = {Association for Computational Linguistics}, address = {Vancouver, Canada}, abstract = {We present a feature-rich knowledge tracing method that captures a student's acquisition and retention of knowledge during a foreign language phrase learning task. We model the student's behavior as making predictions under a log-linear model, and adopt a neural gating mechanism to model how the student updates their log-linear parameters in response to feedback. The gating mechanism allows the model to learn complex patterns of retention and acquisition for each feature, while the log-linear parameterization results in an interpretable knowledge state. We collect human data and evaluate several versions of the model.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We present a feature-rich knowledge tracing method that captures a student's acquisition and retention of knowledge during a foreign language phrase learning task. We model the student's behavior as making predictions under a log-linear model, and adopt a neural gating mechanism to model how the student updates their log-linear parameters in response to feedback. The gating mechanism allows the model to learn complex patterns of retention and acquisition for each feature, while the log-linear parameterization results in an interpretable knowledge state. We collect human data and evaluate several versions of the model. |
Dredze, Mark; Wood-Doughty, Zachary; Quinn, Sandra Crouse; Broniatowski, David A Vaccine opponents' use of Twitter during the 2016 US presidential election: Implications for practice and policy Journal Article Vaccine, 2017. @article{Dredze:2017fv, title = {Vaccine opponents' use of Twitter during the 2016 US presidential election: Implications for practice and policy}, author = {Mark Dredze and Zachary Wood-Doughty and Sandra Crouse Quinn and David A Broniatowski}, year = {2017}, date = {2017-07-01}, journal = {Vaccine}, abstract = {The recent inauguration of President Trump carries with it many public health policy implications. During the election, President Trump, like all political candidates, made policy commitments to various interest groups including vaccine skeptics. These groups celebrated the announcement that Robert Kennedy Jr., a noted proponent of a causal link between vaccines and autism, may chair a commission on vaccines. Furthermore, during the GOP primaries, Mr. Trump endorsed messages associated with vaccine refusal on Twitter, and met with prominent vaccine refusal advocates including Andrew Wakefield, who published the retracted and discredited 1998 Lancet article claiming to link autism to MMR vaccination. In this paper, we show that the new administration has mobilized vaccine refusal advocates, potentially enabling them to influence the national agenda in a manner that could lead to changes in existing vaccination policy.}, keywords = {}, pubstate = {published}, tppubtype = {article} } The recent inauguration of President Trump carries with it many public health policy implications. During the election, President Trump, like all political candidates, made policy commitments to various interest groups including vaccine skeptics. These groups celebrated the announcement that Robert Kennedy Jr., a noted proponent of a causal link between vaccines and autism, may chair a commission on vaccines. Furthermore, during the GOP primaries, Mr. Trump endorsed messages associated with vaccine refusal on Twitter, and met with prominent vaccine refusal advocates including Andrew Wakefield, who published the retracted and discredited 1998 Lancet article claiming to link autism to MMR vaccination. In this paper, we show that the new administration has mobilized vaccine refusal advocates, potentially enabling them to influence the national agenda in a manner that could lead to changes in existing vaccination policy. |
Sakaguchi, Keisuke; Post, Matt; Durme, Benjamin Van Error-repair Dependency Parsing for Ungrammatical Texts Inproceedings Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pp. 189–195, Association for Computational Linguistics, Vancouver, Canada, 2017. @inproceedings{sakaguchi-post-vandurme:2017:Short, title = {Error-repair Dependency Parsing for Ungrammatical Texts}, author = {Keisuke Sakaguchi and Matt Post and Benjamin Van Durme}, url = {http://aclweb.org/anthology/P17-2030}, year = {2017}, date = {2017-07-01}, booktitle = {Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)}, pages = {189--195}, publisher = {Association for Computational Linguistics}, address = {Vancouver, Canada}, abstract = {We propose a new dependency parsing scheme which jointly parses a sentence and repairs grammatical errors by extending the non-directional transition-based formalism of Goldberg and Elhadad (2010) with three additional actions: SUBSTITUTE, DELETE, INSERT. Because these actions may cause an infinite loop in derivation, we also introduce simple constraints that ensure the parser termination. We evaluate our model with respect to dependency accuracy and grammaticality improvements for ungrammatical sentences, demonstrating the robustness and applicability of our scheme.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We propose a new dependency parsing scheme which jointly parses a sentence and repairs grammatical errors by extending the non-directional transition-based formalism of Goldberg and Elhadad (2010) with three additional actions: SUBSTITUTE, DELETE, INSERT. Because these actions may cause an infinite loop in derivation, we also introduce simple constraints that ensure the parser termination. We evaluate our model with respect to dependency accuracy and grammaticality improvements for ungrammatical sentences, demonstrating the robustness and applicability of our scheme. |
Napoles, Courtney; Sakaguchi, Keisuke; Tetreault, Joel JFLEG: A Fluency Corpus and Benchmark for Grammatical Error Correction Inproceedings Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pp. 229–234, Association for Computational Linguistics, Valencia, Spain, 2017. @inproceedings{napoles-sakaguchi-tetreault:2017:EACLshort, title = {JFLEG: A Fluency Corpus and Benchmark for Grammatical Error Correction}, author = {Courtney Napoles and Keisuke Sakaguchi and Joel Tetreault}, url = {http://www.aclweb.org/anthology/E17-2037}, year = {2017}, date = {2017-04-01}, booktitle = {Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers}, pages = {229--234}, publisher = {Association for Computational Linguistics}, address = {Valencia, Spain}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Vieira, Tim; Francis-Landau, Matthew; Filardo, Nathaniel Wesley; Khorasani, Farzad; Eisner, Jason Dyna: Toward a Self-Optimizing Declarative Language for Machine Learning Applications Inproceedings Proceedings of the First ACM SIGPLAN Workshop on Machine Learning and Programming Languages (MAPL), pp. 8–17, Barcelona, 2017, ISBN: 978-1-4503-5071-6. @inproceedings{vieira-et-al-2017, title = {Dyna: Toward a Self-Optimizing Declarative Language for Machine Learning Applications}, author = {Tim Vieira and Matthew Francis-Landau and Nathaniel Wesley Filardo and Farzad Khorasani and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#vieira-et-al-2017}, isbn = {978-1-4503-5071-6}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the First ACM SIGPLAN Workshop on Machine Learning and Programming Languages (MAPL)}, pages = {8--17}, address = {Barcelona}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Cotterell, Ryan; Poliak, Adam; Durme, Benjamin Van; Eisner, Jason Explaining and Generalizing Skip-Gram through Exponential Family Principal Component Analysis Inproceedings Proceedings of the Conference of the European Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 175–181, Valencia, Spain, 2017. @inproceedings{cotterell-et-al-2017-eaclb, title = {Explaining and Generalizing Skip-Gram through Exponential Family Principal Component Analysis}, author = {Ryan Cotterell and Adam Poliak and Benjamin Van Durme and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#cotterell-et-al-2017-eacl}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Conference of the European Chapter of the Association for Computational Linguistics: Human Language Technologies}, pages = {175--181}, address = {Valencia, Spain}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Wang, Dingquan; Eisner, Jason Fine-Grained Prediction of Syntactic Typology: Discovering Latent Structure with em Supervised Learning Journal Article Transactions of the Association for Computational Linguistics (TACL), 5 , pp. 147–161, 2017, ISSN: 2307-387X. @article{wang-eisner-2017, title = {Fine-Grained Prediction of Syntactic Typology: Discovering Latent Structure with em Supervised Learning}, author = {Dingquan Wang and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#wang-eisner-2017}, issn = {2307-387X}, year = {2017}, date = {2017-01-01}, journal = {Transactions of the Association for Computational Linguistics (TACL)}, volume = {5}, pages = {147--161}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Vieira, Tim; Eisner, Jason Learning to Prune: Exploring the Frontier of Fast and Accurate Parsing Journal Article Transactions of the Association for Computational Linguistics (TACL), 5 , pp. 263–278, 2017, ISSN: 2307-387X. @article{vieira-eisner-2017, title = {Learning to Prune: Exploring the Frontier of Fast and Accurate Parsing}, author = {Tim Vieira and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#vieira-eisner-2017}, issn = {2307-387X}, year = {2017}, date = {2017-01-01}, journal = {Transactions of the Association for Computational Linguistics (TACL)}, volume = {5}, pages = {263--278}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Wang, Yiming; Peddinti, Vijayaditya; Xu, Hainan; Zhang, Xiaohui; Povey, Daniel; Khudanpur, Sanjeev Backstitch: Counteracting Finite-sample Bias via Negative Steps Journal Article Interspeech (submitted), 2017. @article{wang2017backstitch, title = {Backstitch: Counteracting Finite-sample Bias via Negative Steps}, author = {Yiming Wang and Vijayaditya Peddinti and Hainan Xu and Xiaohui Zhang and Daniel Povey and Sanjeev Khudanpur}, year = {2017}, date = {2017-01-01}, journal = {Interspeech (submitted)}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Garcia-Romero, Daniel; Snyder, David; Sell, Gregory; Povey, Daniel; McCree, Alan Speaker diarization using deep neural network embeddings Inproceedings 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, pp. 4930–4934, 2017. @inproceedings{garcia2017speaker, title = {Speaker diarization using deep neural network embeddings}, author = {Daniel Garcia-Romero and David Snyder and Gregory Sell and Daniel Povey and Alan McCree}, year = {2017}, date = {2017-01-01}, booktitle = {2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE}, pages = {4930--4934}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Peddinti, Vijayaditya; Wang, Yiming; Povey, Daniel; Khudanpur, Sanjeev Low latency modeling of temporal contexts Journal Article IEEE Signal Processing Letters (submitted), 2017. @article{peddinti2017low, title = {Low latency modeling of temporal contexts}, author = {Vijayaditya Peddinti and Yiming Wang and Daniel Povey and Sanjeev Khudanpur}, year = {2017}, date = {2017-01-01}, journal = {IEEE Signal Processing Letters (submitted)}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Cheng, Gaofeng; Peddinti, Vijayaditya; Povey, Daniel; Manohar, Vimal; Khudanpur, Sanjeev; Yan, Yonghong An exploration of dropout with lstms Inproceedings Proceedings of Interspeech, 2017. @inproceedings{cheng2017exploration, title = {An exploration of dropout with lstms}, author = {Gaofeng Cheng and Vijayaditya Peddinti and Daniel Povey and Vimal Manohar and Sanjeev Khudanpur and Yonghong Yan}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of Interspeech}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Ko, Tom; Peddinti, Vijayaditya; Povey, Daniel; Seltzer, Michael L; Khudanpur, Sanjeev A study on data augmentation of reverberant speech for robust speech recognition Inproceedings Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference on, pp. 5220–5224, IEEE 2017. @inproceedings{ko2017study, title = {A study on data augmentation of reverberant speech for robust speech recognition}, author = {Tom Ko and Vijayaditya Peddinti and Daniel Povey and Michael L Seltzer and Sanjeev Khudanpur}, year = {2017}, date = {2017-01-01}, booktitle = {Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference on}, pages = {5220--5224}, organization = {IEEE}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Ayers, John W; Althouse, Benjamin; Leas, Eric C; Dredze, Mark; Allem, Jon-Patrick Internet searches for suicide following the release of 13 Reasons Why Journal Article JAMA Internal Medicine, 2017. @article{doi:10.1001/jamainternmed.2017.3333, title = {Internet searches for suicide following the release of 13 Reasons Why}, author = {John W Ayers and Benjamin Althouse and Eric C Leas and Mark Dredze and Jon-Patrick Allem}, doi = {10.1001/jamainternmed.2017.3333}, year = {2017}, date = {2017-01-01}, journal = {JAMA Internal Medicine}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Andy, Anietie; Dredze, Mark; Rwebangira, Mugizi; Callison-Burch, Chris Constructing an Alias List for Named Entities during an Event Inproceedings EMNLP Workshop on Noisy User-generated Text (W-NUT), 2017. @inproceedings{Andy:2017rr, title = {Constructing an Alias List for Named Entities during an Event}, author = {Anietie Andy and Mark Dredze and Mugizi Rwebangira and Chris Callison-Burch}, year = {2017}, date = {2017-01-01}, booktitle = {EMNLP Workshop on Noisy User-generated Text (W-NUT)}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Peng, Nanyun; Dredze, Mark Multi-task Domain Adaptation for Sequence Tagging Inproceedings ACL Workshop on Representation Learning for NLP (RepL4NLP), 2017. @inproceedings{Peng:2017ye, title = {Multi-task Domain Adaptation for Sequence Tagging}, author = {Nanyun Peng and Mark Dredze}, year = {2017}, date = {2017-01-01}, booktitle = {ACL Workshop on Representation Learning for NLP (RepL4NLP)}, abstract = {Many domain adaptation approaches rely on learning cross domain shared representations to transfer the knowledge learned in one domain to other domains. Traditional domain adaptation only considers adapting for one task. In this paper, we explore multi-task representation learning under the domain adaptation scenario. We propose a neural network framework that supports domain adaptation for multiple tasks simultaneously, and learns shared representations that better generalize for domain adaptation. We apply the proposed framework to domain adaptation for sequence tagging problems considering two tasks: Chinese word segmentation and named entity recognition. Experiments show that multi-task domain adaptation works better than disjoint domain adaptation for each task, and achieves the state-of-the-art results for both tasks in the social media domain.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Many domain adaptation approaches rely on learning cross domain shared representations to transfer the knowledge learned in one domain to other domains. Traditional domain adaptation only considers adapting for one task. In this paper, we explore multi-task representation learning under the domain adaptation scenario. We propose a neural network framework that supports domain adaptation for multiple tasks simultaneously, and learns shared representations that better generalize for domain adaptation. We apply the proposed framework to domain adaptation for sequence tagging problems considering two tasks: Chinese word segmentation and named entity recognition. Experiments show that multi-task domain adaptation works better than disjoint domain adaptation for each task, and achieves the state-of-the-art results for both tasks in the social media domain. |
Wood-Doughty, Zach; Smith, Michael; Broniatowski, David; Dredze, Mark How Does Twitter User Behavior Vary Across Demographic Groups? Inproceedings ACL Workshop on Natural Language Processing and Computational Social Science, 2017. @inproceedings{Wood-Doughty:2017lr, title = {How Does Twitter User Behavior Vary Across Demographic Groups?}, author = {Zach Wood-Doughty and Michael Smith and David Broniatowski and Mark Dredze}, year = {2017}, date = {2017-01-01}, booktitle = {ACL Workshop on Natural Language Processing and Computational Social Science}, abstract = {Demographically-tagged social media messages are a common source of data for computational social science. While these messages can indicate differences in beliefs and behaviors between demographic groups, we do not have a clear understanding of how different demographic groups use platforms such as Twitter. This paper presents a preliminary analysis of how groups' differing behaviors may confound analyses of the groups themselves. We analyzed one million Twitter users by first inferring demographic attributes, and then measuring several indicators of Twitter behavior. We find differences in these indicators across demographic groups, suggesting that there may be underlying differences in how different demographic groups use Twitter.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Demographically-tagged social media messages are a common source of data for computational social science. While these messages can indicate differences in beliefs and behaviors between demographic groups, we do not have a clear understanding of how different demographic groups use platforms such as Twitter. This paper presents a preliminary analysis of how groups' differing behaviors may confound analyses of the groups themselves. We analyzed one million Twitter users by first inferring demographic attributes, and then measuring several indicators of Twitter behavior. We find differences in these indicators across demographic groups, suggesting that there may be underlying differences in how different demographic groups use Twitter. |
Allem, Jon-Patrick; Leas, Eric C; Caputi, Theodore L; Dredze, Mark; Althouse, Benjamin M; Noar, Seth M; Ayers, John W The Charlie Sheen Effect on Rapid In-home Human Immunodeficiency Virus Test Sales Journal Article Prevention Science, 2017. @article{Allem:2017qd, title = {The Charlie Sheen Effect on Rapid In-home Human Immunodeficiency Virus Test Sales}, author = {Jon-Patrick Allem and Eric C Leas and Theodore L Caputi and Mark Dredze and Benjamin M Althouse and Seth M Noar and John W Ayers}, year = {2017}, date = {2017-01-01}, journal = {Prevention Science}, abstract = {One in eight of the 1.2 million Americans living with human immunodeficiency virus (HIV) are unaware of their positive status, and untested individuals are responsible for most new infections. As a result, testing is the most cost-effective HIV prevention strategy and must be accelerated when opportunities are presented. Web searches for HIV spiked around actor Charlie Sheen's HIV-positive disclosure. However, it is unknown whether Sheen's disclosure impacted offline behaviors like HIV testing. The goal of this study was to determine if Sheen's HIV disclosure was a record-setting HIV prevention event and determine if Web searches presage increases in testing allowing for rapid detection and reaction in the future. Sales of OraQuick rapid in-home HIV test kits in the USA were monitored weekly from April 12, 2014, to April 16, 2016, alongside Web searches including the terms ``test,'' ``tests,'' or ``testing'' and ``HIV'' as accessed from Google Trends. Changes in OraQuick sales around Sheen's disclosure and prediction models using Web searches were assessed. OraQuick sales rose 95% (95% CI, 75--117; p < 0.001) of the week of Sheen's disclosure and remained elevated for 4 more weeks (p < 0.05). In total, there were 8225 more sales than expected around Sheen's disclosure, surpassing World AIDS Day by a factor of about 7. Moreover, Web searches mirrored OraQuick sales trends (r = 0.79), demonstrating their ability to presage increases in testing. The ``Charlie Sheen effect'' represents an important opportunity for a public health response, and in the future, Web searches can be used to detect and act on more opportunities to foster prevention behaviors.}, keywords = {}, pubstate = {published}, tppubtype = {article} } One in eight of the 1.2 million Americans living with human immunodeficiency virus (HIV) are unaware of their positive status, and untested individuals are responsible for most new infections. As a result, testing is the most cost-effective HIV prevention strategy and must be accelerated when opportunities are presented. Web searches for HIV spiked around actor Charlie Sheen's HIV-positive disclosure. However, it is unknown whether Sheen's disclosure impacted offline behaviors like HIV testing. The goal of this study was to determine if Sheen's HIV disclosure was a record-setting HIV prevention event and determine if Web searches presage increases in testing allowing for rapid detection and reaction in the future. Sales of OraQuick rapid in-home HIV test kits in the USA were monitored weekly from April 12, 2014, to April 16, 2016, alongside Web searches including the terms ``test,'' ``tests,'' or ``testing'' and ``HIV'' as accessed from Google Trends. Changes in OraQuick sales around Sheen's disclosure and prediction models using Web searches were assessed. OraQuick sales rose 95% (95% CI, 75--117; p < 0.001) of the week of Sheen's disclosure and remained elevated for 4 more weeks (p < 0.05). In total, there were 8225 more sales than expected around Sheen's disclosure, surpassing World AIDS Day by a factor of about 7. Moreover, Web searches mirrored OraQuick sales trends (r = 0.79), demonstrating their ability to presage increases in testing. The ``Charlie Sheen effect'' represents an important opportunity for a public health response, and in the future, Web searches can be used to detect and act on more opportunities to foster prevention behaviors. |
Gao, Ning; Oard, Douglas; Dredze, Mark Support for Interactive Identification of Mentioned Entities in Conversational Speech Inproceedings International Conference on Research and Development in Information Retrieval (SIGIR) (short paper), 2017. @inproceedings{Gao:2017wo, title = {Support for Interactive Identification of Mentioned Entities in Conversational Speech}, author = {Ning Gao and Douglas Oard and Mark Dredze}, year = {2017}, date = {2017-01-01}, booktitle = {International Conference on Research and Development in Information Retrieval (SIGIR) (short paper)}, abstract = {Searching conversational speech poses several new challenges, among which is how the searcher will make sense of what they find. This paper describes our initial experiments with a freely available collection of Enron telephone conversations. Our goal is to help the user make sense of search results by finding information about mentioned people, places and organizations. Because automated entity recognition is not yet sufficiently accurate on conversational telephone speech, we ask the user to transcribe just the name, and to indicate where in the recording it was heard. We then seek to link that mention to other mentions of the same entity in a variety of sources (in our experiments, in email and in Wikipedia). We cast this as an entity linking problem, and achieve promising results by utilizing social network features to help compensate for the limited accuracy of automatic transcription for this challenging content.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Searching conversational speech poses several new challenges, among which is how the searcher will make sense of what they find. This paper describes our initial experiments with a freely available collection of Enron telephone conversations. Our goal is to help the user make sense of search results by finding information about mentioned people, places and organizations. Because automated entity recognition is not yet sufficiently accurate on conversational telephone speech, we ask the user to transcribe just the name, and to indicate where in the recording it was heard. We then seek to link that mention to other mentions of the same entity in a variety of sources (in our experiments, in email and in Wikipedia). We cast this as an entity linking problem, and achieve promising results by utilizing social network features to help compensate for the limited accuracy of automatic transcription for this challenging content. |
Wolfe, Travis; Dredze, Mark; Durme, Benjamin Van Pocket Knowledge Base Population Inproceedings Association for Computational Linguistics (ACL) (short paper), 2017. @inproceedings{Wolfe:2017dq, title = {Pocket Knowledge Base Population}, author = {Travis Wolfe and Mark Dredze and Benjamin Van Durme}, year = {2017}, date = {2017-01-01}, booktitle = {Association for Computational Linguistics (ACL) (short paper)}, abstract = {Existing Knowledge Base Population methods extract relations from a closed relational schema with limited coverage, leading to sparse KBs. We propose Pocket Knowledge Base Population (PKBP), the task of dynamically constructing a KB of entities related to a query and finding the best characterization of relationships between entities. We describe novel Open Information Extraction methods which leverage the PKB to find informative trigger words. We evaluate using existing KBP shared-task data as well as new annotations collected for this work. Our methods produce high quality KBs from just text with many more entities and relationships than existing KBP systems.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Existing Knowledge Base Population methods extract relations from a closed relational schema with limited coverage, leading to sparse KBs. We propose Pocket Knowledge Base Population (PKBP), the task of dynamically constructing a KB of entities related to a query and finding the best characterization of relationships between entities. We describe novel Open Information Extraction methods which leverage the PKB to find informative trigger words. We evaluate using existing KBP shared-task data as well as new annotations collected for this work. Our methods produce high quality KBs from just text with many more entities and relationships than existing KBP systems. |
Gao, Ning; Dredze, Mark; Oard, Douglas Person Entity Linking in Email with NIL Detection Journal Article Journal of the Association for Information Science and Technology (JAIST), 2017. @article{Gao:2017pb, title = {Person Entity Linking in Email with NIL Detection}, author = {Ning Gao and Mark Dredze and Douglas Oard}, year = {2017}, date = {2017-01-01}, journal = {Journal of the Association for Information Science and Technology (JAIST)}, abstract = {For each specific mention of an entity found in a text, the goal of entity linking is to determine whether the referenced entity is present in an existing knowledge base, and if so to determine which KB entity is the correct referent. Entity linking has been well explored for dissemination-oriented sources such as news stories, blogs, and microblog posts, but the limited work to date on ``conversational'' sources such as email or text chat has not yet attempted to determine when the referent entity is not in the knowledge base (a task known as ``NIL detection''). This article presents a supervised machine learning system for linking named mentions of people in email messages to a collection-specific knowledge base, and that is also capable of NIL detection. This system learns from manually annotated training examples to leverage a rich set of features. The entity linking accuracy for entities present in the knowledge base is substantially and significantly better than the best previously reported results on the Enron email collection, comparable accuracy is reported for the challenging NIL detection task, and these results are for the first time replicated on a second email collection from a different source with comparable results.}, keywords = {}, pubstate = {published}, tppubtype = {article} } For each specific mention of an entity found in a text, the goal of entity linking is to determine whether the referenced entity is present in an existing knowledge base, and if so to determine which KB entity is the correct referent. Entity linking has been well explored for dissemination-oriented sources such as news stories, blogs, and microblog posts, but the limited work to date on ``conversational'' sources such as email or text chat has not yet attempted to determine when the referent entity is not in the knowledge base (a task known as ``NIL detection''). This article presents a supervised machine learning system for linking named mentions of people in email messages to a collection-specific knowledge base, and that is also capable of NIL detection. This system learns from manually annotated training examples to leverage a rich set of features. The entity linking accuracy for entities present in the knowledge base is substantially and significantly better than the best previously reported results on the Enron email collection, comparable accuracy is reported for the challenging NIL detection task, and these results are for the first time replicated on a second email collection from a different source with comparable results. |
Irvine, Ann; Dredze, Mark Harmonic Grammar, Optimality Theory, and Syntax Learnability: An Empirical Exploration of Czech Word Order Unpublished 2017. @unpublished{1702.05793, title = {Harmonic Grammar, Optimality Theory, and Syntax Learnability: An Empirical Exploration of Czech Word Order}, author = {Ann Irvine and Mark Dredze}, year = {2017}, date = {2017-01-01}, abstract = {This work presents a systematic theoretical and empirical comparison of the major algorithms that have been proposed for learning Harmonic and Optimality Theory grammars (HG and OT, respectively). By comparing learning algorithms, we are also able to compare the closely related OT and HG frameworks themselves. Experimental results show that the additional expressivity of the HG framework over OT affords performance gains in the task of predicting the surface word order of Czech sentences. We compare the perceptron with the classic Gradual Learning Algorithm (GLA), which learns OT grammars, as well as the popular Maximum Entropy model. In addition to showing that the perceptron is theoretically appealing, our work shows that the performance of the HG model it learns approaches that of the upper bound in prediction accuracy on a held out test set and that it is capable of accurately modeling observed variation.}, keywords = {}, pubstate = {published}, tppubtype = {unpublished} } This work presents a systematic theoretical and empirical comparison of the major algorithms that have been proposed for learning Harmonic and Optimality Theory grammars (HG and OT, respectively). By comparing learning algorithms, we are also able to compare the closely related OT and HG frameworks themselves. Experimental results show that the additional expressivity of the HG framework over OT affords performance gains in the task of predicting the surface word order of Czech sentences. We compare the perceptron with the classic Gradual Learning Algorithm (GLA), which learns OT grammars, as well as the popular Maximum Entropy model. In addition to showing that the perceptron is theoretically appealing, our work shows that the performance of the HG model it learns approaches that of the upper bound in prediction accuracy on a held out test set and that it is capable of accurately modeling observed variation. |
Benton, Adrian; Coppersmith, Glen; Dredze, Mark Ethical Research Protocols for Social Media Health Research Inproceedings EACL Workshop on Ethics in Natural Language Processing, 2017. @inproceedings{Benton:2017lq, title = {Ethical Research Protocols for Social Media Health Research}, author = {Adrian Benton and Glen Coppersmith and Mark Dredze}, year = {2017}, date = {2017-01-01}, booktitle = {EACL Workshop on Ethics in Natural Language Processing}, abstract = {Social media have transformed data driven research in political science, the social sciences, health, and medicine. Since health research often touches on sensitive topics that relate to ethics of treatment and patient privacy, similar ethical considerations should be acknowledged when using social media data in health research. While much has been said regarding the ethical considerations of social media research, health research leads to an additional set of concerns. We provide practical suggestions in the form of guidelines for researchers working with social media data in health research. These guidelines can inform an IRB proposal for researchers new to social media health research.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Social media have transformed data driven research in political science, the social sciences, health, and medicine. Since health research often touches on sensitive topics that relate to ethics of treatment and patient privacy, similar ethical considerations should be acknowledged when using social media data in health research. While much has been said regarding the ethical considerations of social media research, health research leads to an additional set of concerns. We provide practical suggestions in the form of guidelines for researchers working with social media data in health research. These guidelines can inform an IRB proposal for researchers new to social media health research. |
Ayers, John; Leas, Eric C; Allem, Jon-Patrick; Benton, Adrian; Dredze, Mark; Althouse, Benjamin M; Cruz, Tess B; Unger, Jennifer B Why Do People Use Electronic Nicotine Delivery Systems (Electronic Cigarettes)? A Content Analysis of Twitter, 2012-2015 Journal Article PLoS One, 2017. @article{Ayers:2017qf, title = {Why Do People Use Electronic Nicotine Delivery Systems (Electronic Cigarettes)? A Content Analysis of Twitter, 2012-2015}, author = {John Ayers and Eric C Leas and Jon-Patrick Allem and Adrian Benton and Mark Dredze and Benjamin M Althouse and Tess B Cruz and Jennifer B Unger}, year = {2017}, date = {2017-01-01}, journal = {PLoS One}, abstract = {The reasons for using electronic nicotine delivery systems (ENDS) are poorly understood and are primarily documented by expensive cross-sectional surveys that use preconceived close-ended response options rather than allowing respondents to use their own words. We passively identify the reasons for using ENDS longitudinally from a content analysis of public postings on Twitter. All English language public tweets including several ENDS terms (e.g., ``e-cigarette'' or ``vape'') were captured from the Twitter data stream during 2012 and 2015. After excluding spam, advertisements, and retweets, posts indicating a rationale for vaping were retained. The specific reasons for vaping were then inferred based on a supervised content analysis using annotators from Amazon's Mechanical Turk. During 2012 quitting combustibles was the most cited reason for using ENDS with 43% (95%CI 39--48) of all reason-related tweets cited quitting combustibles, e.g., ``I couldn't quit till I tried ecigs,'' eclipsing the second most cited reason by more than double. Other frequently cited reasons in 2012 included ENDS's social image (21%; 95%CI 18--25), use indoors (14%; 95%CI 11--17), flavors (14%; 95%CI 11--17), safety relative to combustibles (9%; 95%CI 7--11), cost (3%; 95%CI 2--5) and favorable odor (2%; 95%CI 1--3). By 2015 the reasons for using ENDS cited on Twitter had shifted. Both quitting combustibles and use indoors significantly declined in mentions to 29% (95%CI 24--33) and 12% (95%CI 9--16), respectively. At the same time, social image increased to 37% (95%CI 32--43) and lack of odor increased to 5% (95%CI 2--5), the former leading all cited reasons in 2015. Our data suggest the reasons people vape are shifting away from cessation and toward social image. The data also show how the ENDS market is responsive to a changing policy landscape. For instance, smoking indoors was less frequently cited in 2015 as indoor smoking restrictions became more common. Because the data and analytic approach are scalable, adoption of our strategies in the field can inform follow-up survey-based surveillance (so the right questions are asked), interventions, and policies for ENDS.}, keywords = {}, pubstate = {published}, tppubtype = {article} } The reasons for using electronic nicotine delivery systems (ENDS) are poorly understood and are primarily documented by expensive cross-sectional surveys that use preconceived close-ended response options rather than allowing respondents to use their own words. We passively identify the reasons for using ENDS longitudinally from a content analysis of public postings on Twitter. All English language public tweets including several ENDS terms (e.g., ``e-cigarette'' or ``vape'') were captured from the Twitter data stream during 2012 and 2015. After excluding spam, advertisements, and retweets, posts indicating a rationale for vaping were retained. The specific reasons for vaping were then inferred based on a supervised content analysis using annotators from Amazon's Mechanical Turk. During 2012 quitting combustibles was the most cited reason for using ENDS with 43% (95%CI 39--48) of all reason-related tweets cited quitting combustibles, e.g., ``I couldn't quit till I tried ecigs,'' eclipsing the second most cited reason by more than double. Other frequently cited reasons in 2012 included ENDS's social image (21%; 95%CI 18--25), use indoors (14%; 95%CI 11--17), flavors (14%; 95%CI 11--17), safety relative to combustibles (9%; 95%CI 7--11), cost (3%; 95%CI 2--5) and favorable odor (2%; 95%CI 1--3). By 2015 the reasons for using ENDS cited on Twitter had shifted. Both quitting combustibles and use indoors significantly declined in mentions to 29% (95%CI 24--33) and 12% (95%CI 9--16), respectively. At the same time, social image increased to 37% (95%CI 32--43) and lack of odor increased to 5% (95%CI 2--5), the former leading all cited reasons in 2015. Our data suggest the reasons people vape are shifting away from cessation and toward social image. The data also show how the ENDS market is responsive to a changing policy landscape. For instance, smoking indoors was less frequently cited in 2015 as indoor smoking restrictions became more common. Because the data and analytic approach are scalable, adoption of our strategies in the field can inform follow-up survey-based surveillance (so the right questions are asked), interventions, and policies for ENDS. |
Nastasi, Anthony; Bryant, Tyler; Canner, Joseph K; Dredze, Mark; Camp, Melissa S; Nagarajan, Neeraja Breast Cancer Screening and Social Media: a Content Analysis of Evidence Use and Guideline Opinions on Twitter Journal Article Journal of Cancer Education, 2017. @article{Nastasi:2017qq, title = {Breast Cancer Screening and Social Media: a Content Analysis of Evidence Use and Guideline Opinions on Twitter}, author = {Anthony Nastasi and Tyler Bryant and Joseph K Canner and Mark Dredze and Melissa S Camp and Neeraja Nagarajan}, year = {2017}, date = {2017-01-01}, journal = {Journal of Cancer Education}, abstract = {There is ongoing debate regarding the best mammography screening practices. Twitter has become a powerful tool for disseminating medical news and fostering healthcare conversations; however, little work has been done examining these conversations in the context of how users are sharing evidence and discussing current guidelines for breast cancer screening. To characterize the Twitter conversation on mammography and assess the quality of evidence used as well as opinions regarding current screening guidelines, individual tweets using mammography-related hashtags were prospectively pulled from Twitter from 5 November 2015 to 11 December 2015. Content analysis was performed on the tweets by abstracting data related to user demographics, content, evidence use, and guideline opinions. Standard descriptive statistics were used to summarize the results. Comparisons were made by demographics, tweet type (testable claim, advice, personal experience, etc.), and user type (non-healthcare, physician, cancer specialist, etc.). The primary outcomes were how users are tweeting about breast cancer screening, the quality of evidence they are using, and their opinions regarding guidelines. The most frequent user type of the 1345 tweets was ``non-healthcare'' with 323 tweets (32.5%). Physicians had 1.87 times higher odds (95% CI, 0.69--5.07) of providing explicit support with a reference and 11.70 times higher odds (95% CI, 3.41--40.13) of posting a tweet likely to be supported by the scientific community compared to non-healthcare users. Only 2.9% of guideline tweets approved of the guidelines while 14.6% claimed to be confused by them. Non-healthcare users comprise a significant proportion of participants in mammography conversations, with tweets often containing claims that are false, not explicitly backed by scientific evidence, and in favor of alternative ``natural'' breast cancer prevention and treatment. Furthermore, users appear to have low approval and confusion regarding screening guidelines. These findings suggest that more efforts are needed to educate and disseminate accurate information to the general public regarding breast cancer prevention modalities, emphasizing the safety of mammography and the harms of replacing conventional prevention and treatment modalities with unsubstantiated alternatives.}, keywords = {}, pubstate = {published}, tppubtype = {article} } There is ongoing debate regarding the best mammography screening practices. Twitter has become a powerful tool for disseminating medical news and fostering healthcare conversations; however, little work has been done examining these conversations in the context of how users are sharing evidence and discussing current guidelines for breast cancer screening. To characterize the Twitter conversation on mammography and assess the quality of evidence used as well as opinions regarding current screening guidelines, individual tweets using mammography-related hashtags were prospectively pulled from Twitter from 5 November 2015 to 11 December 2015. Content analysis was performed on the tweets by abstracting data related to user demographics, content, evidence use, and guideline opinions. Standard descriptive statistics were used to summarize the results. Comparisons were made by demographics, tweet type (testable claim, advice, personal experience, etc.), and user type (non-healthcare, physician, cancer specialist, etc.). The primary outcomes were how users are tweeting about breast cancer screening, the quality of evidence they are using, and their opinions regarding guidelines. The most frequent user type of the 1345 tweets was ``non-healthcare'' with 323 tweets (32.5%). Physicians had 1.87 times higher odds (95% CI, 0.69--5.07) of providing explicit support with a reference and 11.70 times higher odds (95% CI, 3.41--40.13) of posting a tweet likely to be supported by the scientific community compared to non-healthcare users. Only 2.9% of guideline tweets approved of the guidelines while 14.6% claimed to be confused by them. Non-healthcare users comprise a significant proportion of participants in mammography conversations, with tweets often containing claims that are false, not explicitly backed by scientific evidence, and in favor of alternative ``natural'' breast cancer prevention and treatment. Furthermore, users appear to have low approval and confusion regarding screening guidelines. These findings suggest that more efforts are needed to educate and disseminate accurate information to the general public regarding breast cancer prevention modalities, emphasizing the safety of mammography and the harms of replacing conventional prevention and treatment modalities with unsubstantiated alternatives. |
Huang, Xiaolei; Smith, Michael C; Paul, Michael; Ryzhkov, Dmytro; Quinn, Sandra; Broniatowski, David; Dredze, Mark Examining Patterns of Influenza Vaccination in Social Media Inproceedings AAAI Joint Workshop on Health Intelligence (W3PHIAI), 2017. @inproceedings{Huang:2017ygb, title = {Examining Patterns of Influenza Vaccination in Social Media}, author = {Xiaolei Huang and Michael C Smith and Michael Paul and Dmytro Ryzhkov and Sandra Quinn and David Broniatowski and Mark Dredze}, year = {2017}, date = {2017-01-01}, booktitle = {AAAI Joint Workshop on Health Intelligence (W3PHIAI)}, abstract = {Traditional data on influenza vaccination has several limitations: high cost, limited coverage of underrepresented groups, and low sensitivity to emerging public health issues. Social media, such as Twitter, provide an alternative way to understand a population's vaccination-related opinions and behaviors. In this study, we build and employ several natural language classifiers to examine and analyze behavioral patterns regarding influenza vaccination in Twitter across three dimensions: temporality (by week and month), geography (by US region), and demography (by gender). Our best results are highly correlated official government data, with a correlation over 0.90, providing validation of our approach. We then suggest a number of directions for future work.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Traditional data on influenza vaccination has several limitations: high cost, limited coverage of underrepresented groups, and low sensitivity to emerging public health issues. Social media, such as Twitter, provide an alternative way to understand a population's vaccination-related opinions and behaviors. In this study, we build and employ several natural language classifiers to examine and analyze behavioral patterns regarding influenza vaccination in Twitter across three dimensions: temporality (by week and month), geography (by US region), and demography (by gender). Our best results are highly correlated official government data, with a correlation over 0.90, providing validation of our approach. We then suggest a number of directions for future work. |
Nagarajan, Neeraja; Alshaikh, Husain; Nastasi, Anthony; Smart, Blair; Berger, Zackary; Schneider, Eric B; Dredze, Mark; Canner, Joseph K; Ahuja, Nita The Utility of Twitter in Generating High-Quality Conversations about Surgical Care Inproceedings Academic Surgical Congress, 2017. @inproceedings{Nagarajan:2017kx, title = {The Utility of Twitter in Generating High-Quality Conversations about Surgical Care}, author = {Neeraja Nagarajan and Husain Alshaikh and Anthony Nastasi and Blair Smart and Zackary Berger and Eric B Schneider and Mark Dredze and Joseph K Canner and Nita Ahuja}, year = {2017}, date = {2017-01-01}, booktitle = {Academic Surgical Congress}, abstract = {Introduction: There is growing interest among various stakeholders in using social media sites to discuss healthcare issues. However, little is known about how social media sites are used to discuss surgical care. There is also a lack of understanding of the types of content generated and the quality of the information shared in social media platforms about surgical care issues. We therefore sought to identify and summarize conversations on surgical care in Twitter, a popular microblogging website. Methods: A comprehensive list of surgery-related hashtags was used to pull individual tweets from 3/27-4/27/2015. Four independent reviewers blindly analyzed 25 tweets to develop themes for extraction from a larger sample. The themes were broadly divided further to obtain data at the levels of the user, the tweet, the content of the tweet and personal information shared (Figure I). Standard descriptive statistical analysis and simple logistic regression analysis was used. Results: In total, 17,783 tweets were pulled and 1000 from 615 unique users were randomly selected for analysis. Most users were from North America (62.4%) and non-healthcare related individuals (31.8%). Healthcare organizations generated 12.4%, and surgeons 9.5%, of tweets. Overall, 67.4% were original tweets and 79.0% contained a hyperlink (11% to healthcare and 8.7% to peer-reviewed sources). The common areas of surgery discussed were global surgery/health systems (18.4%), followed by general surgery (15.6%). Among personal tweets (n=236), 31.1% concerned surgery on family/friends and 24.4% on the user; 61.1% discussed procedures already performed and 58.0% used positive language about their personal experience with surgical care. Surgical news/opinion was present in 45% of tweets and 13.7% contained evidence-based information. Non-healthcare professionals were 53.5% (95% CI: 3.8%-77.5%}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Introduction: There is growing interest among various stakeholders in using social media sites to discuss healthcare issues. However, little is known about how social media sites are used to discuss surgical care. There is also a lack of understanding of the types of content generated and the quality of the information shared in social media platforms about surgical care issues. We therefore sought to identify and summarize conversations on surgical care in Twitter, a popular microblogging website. Methods: A comprehensive list of surgery-related hashtags was used to pull individual tweets from 3/27-4/27/2015. Four independent reviewers blindly analyzed 25 tweets to develop themes for extraction from a larger sample. The themes were broadly divided further to obtain data at the levels of the user, the tweet, the content of the tweet and personal information shared (Figure I). Standard descriptive statistical analysis and simple logistic regression analysis was used. Results: In total, 17,783 tweets were pulled and 1000 from 615 unique users were randomly selected for analysis. Most users were from North America (62.4%) and non-healthcare related individuals (31.8%). Healthcare organizations generated 12.4%, and surgeons 9.5%, of tweets. Overall, 67.4% were original tweets and 79.0% contained a hyperlink (11% to healthcare and 8.7% to peer-reviewed sources). The common areas of surgery discussed were global surgery/health systems (18.4%), followed by general surgery (15.6%). Among personal tweets (n=236), 31.1% concerned surgery on family/friends and 24.4% on the user; 61.1% discussed procedures already performed and 58.0% used positive language about their personal experience with surgical care. Surgical news/opinion was present in 45% of tweets and 13.7% contained evidence-based information. Non-healthcare professionals were 53.5% (95% CI: 3.8%-77.5% |
Zhang, Xiaohui; Manohar, Vimal; Povey, Daniel; Khudanpur, Sanjeev Acoustic data-driven lexicon learning based on a greedy pronunciation selection framework Journal Article arXiv preprint arXiv:1706.03747, 2017. @article{zhang2017acoustic, title = {Acoustic data-driven lexicon learning based on a greedy pronunciation selection framework}, author = {Xiaohui Zhang and Vimal Manohar and Daniel Povey and Sanjeev Khudanpur}, year = {2017}, date = {2017-01-01}, journal = {arXiv preprint arXiv:1706.03747}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Trmal, Jan; Wiesner, Matthew; Peddinti, Vijayaditya; Zhang, Xiaohui; Ghahremani, Pegah; Wang, Yiming; Manohar, Vimal; Xu, Hainan; Povey, Daniel; Khudanpur, Sanjeev The Kaldi OpenKWS System: Improving Low Resource Keyword Search Journal Article 2017. @article{trmal2017kaldi, title = {The Kaldi OpenKWS System: Improving Low Resource Keyword Search}, author = {Jan Trmal and Matthew Wiesner and Vijayaditya Peddinti and Xiaohui Zhang and Pegah Ghahremani and Yiming Wang and Vimal Manohar and Hainan Xu and Daniel Povey and Sanjeev Khudanpur}, year = {2017}, date = {2017-01-01}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Sakaguchi, Keisuke; Duh, Kevin; Post, Matt; Durme, Benjamin Van Robsut Wrod Reocginiton via Semi-Character Recurrent Neural Network Inproceedings Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, February 4-9, 2017, San Francisco, California, USA., pp. 3281–3287, 2017. @inproceedings{DBLP:conf/aaai/SakaguchiDPD17, title = {Robsut Wrod Reocginiton via Semi-Character Recurrent Neural Network}, author = {Keisuke Sakaguchi and Kevin Duh and Matt Post and Benjamin Van Durme}, url = {http://aaai.org/ocs/index.php/AAAI/AAAI17/paper/view/14332}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, February 4-9, 2017, San Francisco, California, USA.}, pages = {3281--3287}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Nagarajan, Neeraja; Alshaikh, Husain; Nastasi, Anthony; Smart, Blair; Berger, Zackary; Schneider, Eric B; Dredze, Mark; Canner, Joseph K; Ahuja, Nita The Utility of Twitter in Generating High-Quality Conversations about Surgical Care Inproceedings Academic Surgical Congress, 2017. @inproceedings{Nagarajan:2017kxb, title = {The Utility of Twitter in Generating High-Quality Conversations about Surgical Care}, author = {Neeraja Nagarajan and Husain Alshaikh and Anthony Nastasi and Blair Smart and Zackary Berger and Eric B Schneider and Mark Dredze and Joseph K Canner and Nita Ahuja}, year = {2017}, date = {2017-01-01}, booktitle = {Academic Surgical Congress}, abstract = {Introduction: There is growing interest among various stakeholders in using social media sites to discuss healthcare issues. However, little is known about how social media sites are used to discuss surgical care. There is also a lack of understanding of the types of content generated and the quality of the information shared in social media platforms about surgical care issues. We therefore sought to identify and summarize conversations on surgical care in Twitter, a popular microblogging website. Methods: A comprehensive list of surgery-related hashtags was used to pull individual tweets from 3/27-4/27/2015. Four independent reviewers blindly analyzed 25 tweets to develop themes for extraction from a larger sample. The themes were broadly divided further to obtain data at the levels of the user, the tweet, the content of the tweet and personal information shared (Figure I). Standard descriptive statistical analysis and simple logistic regression analysis was used. Results: In total, 17,783 tweets were pulled and 1000 from 615 unique users were randomly selected for analysis. Most users were from North America (62.4%) and non-healthcare related individuals (31.8%). Healthcare organizations generated 12.4%, and surgeons 9.5%, of tweets. Overall, 67.4% were original tweets and 79.0% contained a hyperlink (11% to healthcare and 8.7% to peer-reviewed sources). The common areas of surgery discussed were global surgery/health systems (18.4%), followed by general surgery (15.6%). Among personal tweets (n=236), 31.1% concerned surgery on family/friends and 24.4% on the user; 61.1% discussed procedures already performed and 58.0% used positive language about their personal experience with surgical care. Surgical news/opinion was present in 45% of tweets and 13.7% contained evidence-based information. Non-healthcare professionals were 53.5% (95% CI: 3.8%-77.5%}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Introduction: There is growing interest among various stakeholders in using social media sites to discuss healthcare issues. However, little is known about how social media sites are used to discuss surgical care. There is also a lack of understanding of the types of content generated and the quality of the information shared in social media platforms about surgical care issues. We therefore sought to identify and summarize conversations on surgical care in Twitter, a popular microblogging website. Methods: A comprehensive list of surgery-related hashtags was used to pull individual tweets from 3/27-4/27/2015. Four independent reviewers blindly analyzed 25 tweets to develop themes for extraction from a larger sample. The themes were broadly divided further to obtain data at the levels of the user, the tweet, the content of the tweet and personal information shared (Figure I). Standard descriptive statistical analysis and simple logistic regression analysis was used. Results: In total, 17,783 tweets were pulled and 1000 from 615 unique users were randomly selected for analysis. Most users were from North America (62.4%) and non-healthcare related individuals (31.8%). Healthcare organizations generated 12.4%, and surgeons 9.5%, of tweets. Overall, 67.4% were original tweets and 79.0% contained a hyperlink (11% to healthcare and 8.7% to peer-reviewed sources). The common areas of surgery discussed were global surgery/health systems (18.4%), followed by general surgery (15.6%). Among personal tweets (n=236), 31.1% concerned surgery on family/friends and 24.4% on the user; 61.1% discussed procedures already performed and 58.0% used positive language about their personal experience with surgical care. Surgical news/opinion was present in 45% of tweets and 13.7% contained evidence-based information. Non-healthcare professionals were 53.5% (95% CI: 3.8%-77.5% |
Rudinger, Rachel; May, Chandler; Durme, Benjamin Van Social Bias in Elicited Natural Language Inferences Inproceedings Proceedings of the First ACL Workshop on Ethics in Natural Language Processing, 2017. @inproceedings{rudinger2017, title = {Social Bias in Elicited Natural Language Inferences}, author = {Rachel Rudinger and Chandler May and Benjamin Van Durme}, url = {http://www.ethicsinnlp.org/workshop/pdf/EthNLP09.pdf}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the First ACL Workshop on Ethics in Natural Language Processing}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
May, Chandler; Duh, Kevin; Van Durme, Benjamin ; Lall, Ashwin Streaming Word Embeddings with the Space-Saving Algorithm Unpublished 2017. @unpublished{may2017, title = {Streaming Word Embeddings with the Space-Saving Algorithm}, author = {Chandler May and Kevin Duh and Benjamin {Van Durme} and Ashwin Lall}, url = {https://arxiv.org/abs/1704.07463}, year = {2017}, date = {2017-01-01}, journal = {ArXiv e-prints}, keywords = {}, pubstate = {published}, tppubtype = {unpublished} } |
Cotterell, Ryan; Kirov, Christo; Sylak-Glassman, John; Walther, Géraldine; Vylomova, Ekaterina; Xia, Patrick; Faruqui, Manaal; ü, K; Yarowsky, David; Eisner, Jason; Hulden, Mans CoNLL-SIGMORPHON 2017 Shared Task: Universal Morphological Reinflection in 52 Languages Inproceedings Proceedings of the Conference on Natural Language Learning: CoNLL-SIGMORPHON Shared Task System Descriptions, pp. 1–30, Vancouver, 2017. @inproceedings{cotterell-et-al-2017-shared, title = {CoNLL-SIGMORPHON 2017 Shared Task: Universal Morphological Reinflection in 52 Languages}, author = {Ryan Cotterell and Christo Kirov and John Sylak-Glassman and Géraldine Walther and Ekaterina Vylomova and Patrick Xia and Manaal Faruqui and K ü and David Yarowsky and Jason Eisner and Mans Hulden}, url = {http://cs.jhu.edu/~jason/papers/#cotterell-et-al-2017-shared}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Conference on Natural Language Learning: CoNLL-SIGMORPHON Shared Task System Descriptions}, pages = {1--30}, address = {Vancouver}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Hulden, Mans; Cotterell, Ryan; Eisner, Jason; Faruqui, Manaal; Kirov, Christo; ü, K; Sylak-Glassman, John; Vylomova, Ekaterina; Walther, Géraldine; Xia, Patrick; Yarowsky, David (Ed.) Proceedings of the CoNLL SIGMORPHON 2007 Shared Task: Universal Morphological Reinflection Book Vancouver, 2017, (125 pages). @book{SIGMORPHON-2017-shared, title = {Proceedings of the CoNLL SIGMORPHON 2007 Shared Task: Universal Morphological Reinflection}, editor = {Mans Hulden and Ryan Cotterell and Jason Eisner and Manaal Faruqui and Christo Kirov and K ü and John Sylak-Glassman and Ekaterina Vylomova and Géraldine Walther and Patrick Xia and David Yarowsky}, url = {http://cs.jhu.edu/~jason/papers/#SIGMORPHON-2017-shared}, year = {2017}, date = {2017-01-01}, address = {Vancouver}, note = {125 pages}, keywords = {}, pubstate = {published}, tppubtype = {book} } |
Knowledge Tracing in Sequential Learning of Inflected Vocabulary Inproceedings Proceedings of the Conference on Natural Language Learning (CoNLL), pp. 238–247, Vancouver, 2017. @inproceedings{renduchintala-et-al-2017-conll, title = {Knowledge Tracing in Sequential Learning of Inflected Vocabulary}, url = {http://cs.jhu.edu/~jason/papers/#renduchintala-et-al-2017-conll}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Conference on Natural Language Learning (CoNLL)}, pages = {238--247}, address = {Vancouver}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Mei, Hongyuan; Eisner, Jason The Neural Ħawkes Process: A Neurally Self-Modulating Multivariate Point Process Journal Article Computing Research Repository, 2017, (First arXiv version December 2016.). @article{mei-eisner-2017, title = {The Neural Ħawkes Process: A Neurally Self-Modulating Multivariate Point Process}, author = {Hongyuan Mei and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#mei-eisner-2017}, year = {2017}, date = {2017-01-01}, journal = {Computing Research Repository}, note = {First arXiv version December 2016.}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Cotterell, Ryan; Eisner, Jason Probabilistic Typology: Deep Generative Models of Vowel Inventories Inproceedings Proceedings of the Association for Computational Linguistics, pp. 1182–1192, Vancouver, 2017, (Best Long Paper Award). @inproceedings{cotterell-eisner-2017-acl, title = {Probabilistic Typology: Deep Generative Models of Vowel Inventories}, author = {Ryan Cotterell and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#cotterell-eisner-2017-acl}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Association for Computational Linguistics}, pages = {1182--1192}, address = {Vancouver}, note = {Best Long Paper Award}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Andrews, Nicholas; Dredze, Mark; Durme, Benjamin Van; Eisner, Jason Bayesian Modeling of Lexical Resources for Low-Resource Settings Inproceedings Proceedings of the Association for Computational Linguistics, pp. 1029–1039, Vancouver, 2017. @inproceedings{andrews-et-al-2017, title = {Bayesian Modeling of Lexical Resources for Low-Resource Settings}, author = {Nicholas Andrews and Mark Dredze and Benjamin Van Durme and Jason Eisner}, url = {http://cs.jhu.edu/~jason/papers/#andrews-et-al-2017}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the Association for Computational Linguistics}, pages = {1029--1039}, address = {Vancouver}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
2016 |
Nadejde, Maria; Birch, Alexandra; Koehn, Philipp A Verb Lexicon Model with Source-side Syntactic Context for String-to-Tree Machine Translation Inproceedings Proceedings of the International Workshop on Spoken Language Translation (IWSLT), Seattle, Washington, USA, 2016. @inproceedings{IWSLT-2016-Nadejde, title = {A Verb Lexicon Model with Source-side Syntactic Context for String-to-Tree Machine Translation}, author = {Maria Nadejde and Alexandra Birch and Philipp Koehn}, url = {http://workshop2016.iwslt.org/downloads/IWSLT_2016_paper_10.pdf}, year = {2016}, date = {2016-12-01}, booktitle = {Proceedings of the International Workshop on Spoken Language Translation (IWSLT)}, address = {Seattle, Washington, USA}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
White, Aaron Steven; Reisinger, Drew; Sakaguchi, Keisuke; Vieira, Tim; Zhang, Sheng; Rudinger, Rachel; Rawlins, Kyle; Durme, Benjamin Van Universal Decompositional Semantics on Universal Dependencies Inproceedings Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pp. 1713–1723, Association for Computational Linguistics, Austin, Texas, 2016. @inproceedings{white-EtAl:2016:EMNLP2016, title = {Universal Decompositional Semantics on Universal Dependencies}, author = {Aaron Steven White and Drew Reisinger and Keisuke Sakaguchi and Tim Vieira and Sheng Zhang and Rachel Rudinger and Kyle Rawlins and Benjamin Van Durme}, url = {https://aclweb.org/anthology/D16-1177}, year = {2016}, date = {2016-11-01}, booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing}, pages = {1713--1723}, publisher = {Association for Computational Linguistics}, address = {Austin, Texas}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Napoles, Courtney; Sakaguchi, Keisuke; Tetreault, Joel There's No Comparison: Reference-less Evaluation Metrics in Grammatical Error Correction Inproceedings Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pp. 2109–2115, Association for Computational Linguistics, Austin, Texas, 2016. @inproceedings{napoles-sakaguchi-tetreault:2016:EMNLP2016, title = {There's No Comparison: Reference-less Evaluation Metrics in Grammatical Error Correction}, author = {Courtney Napoles and Keisuke Sakaguchi and Joel Tetreault}, url = {https://aclweb.org/anthology/D16-1228}, year = {2016}, date = {2016-11-01}, booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing}, pages = {2109--2115}, publisher = {Association for Computational Linguistics}, address = {Austin, Texas}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Nagata, Ryo; Sakaguchi, Keisuke Phrase Structure Annotation and Parsing for Learner English Inproceedings Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 1837–1847, Association for Computational Linguistics, Berlin, Germany, 2016. @inproceedings{nagata-sakaguchi:2016:P16-1, title = {Phrase Structure Annotation and Parsing for Learner English}, author = {Ryo Nagata and Keisuke Sakaguchi}, url = {http://www.aclweb.org/anthology/P16-1173}, year = {2016}, date = {2016-08-01}, booktitle = {Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, pages = {1837--1847}, publisher = {Association for Computational Linguistics}, address = {Berlin, Germany}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Renduchintala, Adithya; Knowles, Rebecca; Koehn, Philipp; Eisner, Jason User Modeling in Language Learning with Macaronic Texts Inproceedings Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 1859–1869, Association for Computational Linguistics, Berlin, Germany, 2016. @inproceedings{renduchintala-EtAl:2016:P16-1, title = {User Modeling in Language Learning with Macaronic Texts}, author = {Adithya Renduchintala and Rebecca Knowles and Philipp Koehn and Jason Eisner}, url = {http://www.aclweb.org/anthology/P16-1175}, year = {2016}, date = {2016-08-01}, booktitle = {Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, pages = {1859--1869}, publisher = {Association for Computational Linguistics}, address = {Berlin, Germany}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Renduchintala, Adithya; Knowles, Rebecca; Koehn, Philipp; Eisner, Jason Creating Interactive Macaronic Interfaces for Language Learning Inproceedings Proceedings of ACL-2016 System Demonstrations, pp. 133–138, Association for Computational Linguistics, Berlin, Germany, 2016. @inproceedings{renduchintala-EtAl:2016:P16-4, title = {Creating Interactive Macaronic Interfaces for Language Learning}, author = {Adithya Renduchintala and Rebecca Knowles and Philipp Koehn and Jason Eisner}, url = {http://anthology.aclweb.org/P/P16/P16-4023}, year = {2016}, date = {2016-08-01}, booktitle = {Proceedings of ACL-2016 System Demonstrations}, pages = {133--138}, publisher = {Association for Computational Linguistics}, address = {Berlin, Germany}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Upcoming Seminars
Sep
22
Fri
12:00 pm
Tal Linzen (JHU) “Structure-Sens...
@ Hackerman Hall B17
Tal Linzen (JHU) “Structure-Sens...
@ Hackerman Hall B17
Sep 22 @ 12:00 pm – 1:15 pm
Abstract Neural networks have recently become ubiquitous in natural language processing systems, but we typically have little understanding of specific capabilities of these networks beyond their overall accuracy in an applied task. The present work[...]
Sep
29
Fri
12:00 pm
Satinder Singh (University of Mi...
@ Hackerman Hall B17
Satinder Singh (University of Mi...
@ Hackerman Hall B17
Sep 29 @ 12:00 pm – 1:15 pm
Abstract The success of Deep Learning (DL) on visual perception has led to rapid progress on Reinforcement Learning (RL) tasks with visual inputs. More recently, Deep Learning is showing promise at certain kinds of supervised[...]
