Publications
| 521. | Paul, Michael; Dredze, Mark: Drug Extraction from the Web: Summarizing Drug Experiences with Multi-Dimensional Topic Models. In: North American Chapter of the Association for Computational Linguistics (NAACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Paul:2013, title = {Drug Extraction from the Web: Summarizing Drug Experiences with Multi-Dimensional Topic Models}, author = {Michael Paul and Mark Dredze}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, abstract = {Multi-dimensional latent text models, such as factorial LDA (f-LDA), capture multiple factors of corpora, creating structured output for researchers to better understand the contents of a corpus. We consider such models for clinical research of new recreational drugs and trends, an important application for mining current information for healthcare workers. We use a "three-dimensional" f-LDA variant to jointly model combinations of drug (marijuana, salvia, etc.), aspect (effects, chemistry, etc.) and route of administration (smoking, oral, etc.) Since a purely unsupervised topic model is unlikely to discover these specific factors of interest, we develop a novel method of incorporating prior knowledge by leveraging user generated tags as priors in our model. We demonstrate that this model can be used as an exploratory tool for learning about these drugs from the Web by applying it to the task of extractive summarization. In addition to providing useful output for this important public health task, our prior-enriched model provides a framework for the application of f-LDA to other tasks}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Multi-dimensional latent text models, such as factorial LDA (f-LDA), capture multiple factors of corpora, creating structured output for researchers to better understand the contents of a corpus. We consider such models for clinical research of new recreational drugs and trends, an important application for mining current information for healthcare workers. We use a "three-dimensional" f-LDA variant to jointly model combinations of drug (marijuana, salvia, etc.), aspect (effects, chemistry, etc.) and route of administration (smoking, oral, etc.) Since a purely unsupervised topic model is unlikely to discover these specific factors of interest, we develop a novel method of incorporating prior knowledge by leveraging user generated tags as priors in our model. We demonstrate that this model can be used as an exploratory tool for learning about these drugs from the Web by applying it to the task of extractive summarization. In addition to providing useful output for this important public health task, our prior-enriched model provides a framework for the application of f-LDA to other tasks |
| 522. | Crammer, Koby; Kulesza, Alex; Dredze, Mark: Adaptive Regularization of Weight Vectors. In: Machine Learning, 91 (2), pp. 155-187, 2013. (Type: Journal Article | Abstract | Links | BibTeX) @article{Crammer:2013fk, title = {Adaptive Regularization of Weight Vectors}, author = {Koby Crammer and Alex Kulesza and Mark Dredze}, doi = {10.1007/s10994-013-5327-x}, year = {2013}, date = {2013-01-01}, journal = {Machine Learning}, volume = {91}, number = {2}, pages = {155-187}, abstract = {We present AROW, an online learning algorithm for binary and multiclass problems that combines large margin training, confidence weighting, and the capacity to handle non-separable data. AROW performs adaptive regularization of the prediction function upon seeing each new instance, allowing it to perform especially well in the presence of label noise. We derive mistake bounds for the binary and multiclass settings that are similar in form to the second order perceptron bound. Our bounds do not assume separability. We also relate our algorithm to recent confidence-weighted online learning techniques. Empirical evaluations show that AROW achieves state-of-the-art performance on a wide range of binary and multiclass tasks, as well as robustness in the face of non-separable data.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We present AROW, an online learning algorithm for binary and multiclass problems that combines large margin training, confidence weighting, and the capacity to handle non-separable data. AROW performs adaptive regularization of the prediction function upon seeing each new instance, allowing it to perform especially well in the presence of label noise. We derive mistake bounds for the binary and multiclass settings that are similar in form to the second order perceptron bound. Our bounds do not assume separability. We also relate our algorithm to recent confidence-weighted online learning techniques. Empirical evaluations show that AROW achieves state-of-the-art performance on a wide range of binary and multiclass tasks, as well as robustness in the face of non-separable data. |
| 523. | Wolfe, Travis; Durme, Benjamin Van; Dredze, Mark; Andrews, Nicholas; Beller, Charley; Callison-Burch, Chris; DeYoung, Jay; Snyder, Justin; Weese, Jonathan; Xu, Tan; Yao, Xuchen: PARMA: A Predicate Argument Aligner. In: Association for Computational Linguistics (ACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Wolfe:2013lrb, title = {PARMA: A Predicate Argument Aligner}, author = {Travis Wolfe and Benjamin Van Durme and Mark Dredze and Nicholas Andrews and Charley Beller and Chris Callison-Burch and Jay DeYoung and Justin Snyder and Jonathan Weese and Tan Xu and Xuchen Yao}, year = {2013}, date = {2013-01-01}, booktitle = {Association for Computational Linguistics (ACL)}, abstract = {We introduce PARMA, a system for cross-document, semantic predicate and argument alignment. Our system integrates popular lexical semantic resources into a simple discriminative model. PARMA achieves state of the art results. We suggest that existing efforts have focused on data that is too easy, and we provide a more difficult dataset based on MT translation references which has a lower baseline which we beat by 17% absolute F1.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We introduce PARMA, a system for cross-document, semantic predicate and argument alignment. Our system integrates popular lexical semantic resources into a simple discriminative model. PARMA achieves state of the art results. We suggest that existing efforts have focused on data that is too easy, and we provide a more difficult dataset based on MT translation references which has a lower baseline which we beat by 17% absolute F1. |
| 524. | Parada, Carolina; Dredze, Mark; Sethy, Abhinav; Rastrow, Ariya: Sub-Lexical and Contextual Modeling of Out-of-Vocabulary Words in Speech Recognition. Human Language Technology Center of Excellence, Johns Hopkins University (10), 2013. (Type: Technical Report | Abstract | BibTeX) @techreport{parada_techb, title = {Sub-Lexical and Contextual Modeling of Out-of-Vocabulary Words in Speech Recognition}, author = {Carolina Parada and Mark Dredze and Abhinav Sethy and Ariya Rastrow}, year = {2013}, date = {2013-01-01}, number = {10}, institution = {Human Language Technology Center of Excellence, Johns Hopkins University}, abstract = {Large vocabulary speech recognition systems fail to recognize words beyond their vocabulary, many of which are information rich terms, like named entities or foreign words. Hybrid word/sub-word systems solve this problem by adding sub-word units to large vocabulary word based systems; new words can then be represented by combinations of sub-word units. We present a novel probabilistic model to learn the sub-word lexicon optimized for a given task. We consider the task of Out Of vocabulary (OOV) word detection, which relies on output from a hybrid system. We combine the proposed hybrid system with confidence based metrics to improve OOV detection performance. Previous work address OOV detection as a binary classification task, where each region is independently classified using local information. We propose to treat OOV detection as a sequence labeling problem, and we show that 1) jointly predicting out-of-vocabulary regions, 2) including contextual information from each region, and 3) learning sub-lexical units optimized for this task, leads to substantial improvements with respect to state-of-the-art on an English Broadcast News and MIT Lectures task.}, keywords = {}, pubstate = {published}, tppubtype = {techreport} } Large vocabulary speech recognition systems fail to recognize words beyond their vocabulary, many of which are information rich terms, like named entities or foreign words. Hybrid word/sub-word systems solve this problem by adding sub-word units to large vocabulary word based systems; new words can then be represented by combinations of sub-word units. We present a novel probabilistic model to learn the sub-word lexicon optimized for a given task. We consider the task of Out Of vocabulary (OOV) word detection, which relies on output from a hybrid system. We combine the proposed hybrid system with confidence based metrics to improve OOV detection performance. Previous work address OOV detection as a binary classification task, where each region is independently classified using local information. We propose to treat OOV detection as a sequence labeling problem, and we show that 1) jointly predicting out-of-vocabulary regions, 2) including contextual information from each region, and 3) learning sub-lexical units optimized for this task, leads to substantial improvements with respect to state-of-the-art on an English Broadcast News and MIT Lectures task. |
| 525. | Dredze, Mark; Paul, Michael J; Bergsma, Shane; Tran, Hieu: Carmen: A Twitter Geolocation System with Applications to Public Health. In: AAAI Workshop on Expanding the Boundaries of Health Informatics Using AI (HIAI), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Dredze:2013ab, title = {Carmen: A Twitter Geolocation System with Applications to Public Health}, author = {Mark Dredze and Michael J Paul and Shane Bergsma and Hieu Tran}, year = {2013}, date = {2013-01-01}, booktitle = {AAAI Workshop on Expanding the Boundaries of Health Informatics Using AI (HIAI)}, abstract = {Public health applications using social media often require accurate, broad-coverage location information. However, the standard information provided by social media APIs, such as Twitter, cover a limited number of messages. This paper presents Carmen, a geolocation system that can determine structured location information for messages provided by the Twitter API. Our system utilizes geocoding tools and a combination of automatic and manual alias resolution methods to infer location structures from GPS positions and user-provided profile data. We show that our system is accurate and covers many locations, and we demonstrate its utility for improving influenza surveillance.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Public health applications using social media often require accurate, broad-coverage location information. However, the standard information provided by social media APIs, such as Twitter, cover a limited number of messages. This paper presents Carmen, a geolocation system that can determine structured location information for messages provided by the Twitter API. Our system utilizes geocoding tools and a combination of automatic and manual alias resolution methods to infer location structures from GPS positions and user-provided profile data. We show that our system is accurate and covers many locations, and we demonstrate its utility for improving influenza surveillance. |
| 526. | Paul, Michael J; Wallace, Byron; Dredze, Mark: What Affects Patient (Dis)satisfaction? Analyzing Online Doctor Ratings with a Joint Topic-Sentiment Model. In: AAAI Workshop on Expanding the Boundaries of Health Informatics Using AI (HIAI), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Paul:2013fkb, title = {What Affects Patient (Dis)satisfaction? Analyzing Online Doctor Ratings with a Joint Topic-Sentiment Model}, author = {Michael J Paul and Byron Wallace and Mark Dredze}, year = {2013}, date = {2013-01-01}, booktitle = {AAAI Workshop on Expanding the Boundaries of Health Informatics Using AI (HIAI)}, abstract = {We analyze patient reviews of doctors using a novel probabilistic joint model of aspect and sentiment based on factorial LDA. We leverage this model to exploit a small set of previously annotated reviews to automatically analyze the topics and sentiment latent in over 50,000 online reviews of physicians (and we make this dataset publicly available). The proposed model outperforms baseline models for this task with respect to model perplexity and sentiment classification. We report the most representative words with respect to positive and negative sentiment along three clinical aspects, thus complementing existing qualitative work exploring patient reviews of physicians.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We analyze patient reviews of doctors using a novel probabilistic joint model of aspect and sentiment based on factorial LDA. We leverage this model to exploit a small set of previously annotated reviews to automatically analyze the topics and sentiment latent in over 50,000 online reviews of physicians (and we make this dataset publicly available). The proposed model outperforms baseline models for this task with respect to model perplexity and sentiment classification. We report the most representative words with respect to positive and negative sentiment along three clinical aspects, thus complementing existing qualitative work exploring patient reviews of physicians. |
| 527. | Snyder, Justin; Knowles, Rebecca; Dredze, Mark; Gormley, Matthew R; Wolfe, Travis: Topic Models and Metadata for Visualizing Text Corpora. In: North American Chapter of the Association for Computational Linguistics (NAACL) (Demo Paper), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Snyder:2013lrb, title = {Topic Models and Metadata for Visualizing Text Corpora}, author = {Justin Snyder and Rebecca Knowles and Mark Dredze and Matthew R. Gormley and Travis Wolfe}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL) (Demo Paper)}, abstract = {Effectively exploring and analyzing large text corpora requires visualizations that provide a high level summary. Past work has relied on faceted browsing of document metadata or on natural language processing of document text. In this paper, we present a new web-based tool that integrates topics learned from an unsupervised topic model in a faceted browsing experience. The user can manage topics, filter documents by topic and summarize views with metadata and topic graphs. We report a user study of the usefulness of topics in our tool.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Effectively exploring and analyzing large text corpora requires visualizations that provide a high level summary. Past work has relied on faceted browsing of document metadata or on natural language processing of document text. In this paper, we present a new web-based tool that integrates topics learned from an unsupervised topic model in a faceted browsing experience. The user can manage topics, filter documents by topic and summarize views with metadata and topic graphs. We report a user study of the usefulness of topics in our tool. |
| 528. | Karakos, Damianos; Dredze, Mark; Khudanpur, Sanjeev: Estimating Confusions in the ASR Channel for Improved Topic-based Language Model Adaptation. Johns Hopkins University http://arxiv.org/abs/1303.5148, (8), 2013. (Type: Technical Report | Abstract | BibTeX) @techreport{Karakos:2013fkb, title = {Estimating Confusions in the ASR Channel for Improved Topic-based Language Model Adaptation}, author = {Damianos Karakos and Mark Dredze and Sanjeev Khudanpur}, year = {2013}, date = {2013-01-01}, number = {8}, address = {http://arxiv.org/abs/1303.5148}, institution = {Johns Hopkins University}, abstract = {Human language is a combination of elemental languages/domains/styles that change across and sometimes within discourses. Language models, which play a crucial role in speech recognizers and machine translation systems, are particularly sensitive to such changes, unless some form of adaptation takes place. One approach to speech language model adaptation is self-training, in which a language model's parameters are tuned based on automatically transcribed audio. However, transcription errors can misguide self-training, particularly in challenging settings such as conversational speech. In this work, we propose a model that considers the confusions (errors) of the ASR channel. By modeling the likely confusions in the ASR output instead of using just the 1-best, we improve self-training efficacy by obtaining a more reliable reference transcription estimate. We demonstrate improved topic-based language modeling adaptation results over both 1-best and lattice self-training using our ASR channel confusion estimates on telephone conversations.}, keywords = {}, pubstate = {published}, tppubtype = {techreport} } Human language is a combination of elemental languages/domains/styles that change across and sometimes within discourses. Language models, which play a crucial role in speech recognizers and machine translation systems, are particularly sensitive to such changes, unless some form of adaptation takes place. One approach to speech language model adaptation is self-training, in which a language model's parameters are tuned based on automatically transcribed audio. However, transcription errors can misguide self-training, particularly in challenging settings such as conversational speech. In this work, we propose a model that considers the confusions (errors) of the ASR channel. By modeling the likely confusions in the ASR output instead of using just the 1-best, we improve self-training efficacy by obtaining a more reliable reference transcription estimate. We demonstrate improved topic-based language modeling adaptation results over both 1-best and lattice self-training using our ASR channel confusion estimates on telephone conversations. |
| 529. | Bergsma, Shane; Dredze, Mark; Durme, Benjamin Van; Wilson, Theresa; Yarowsky, David: Broadly Improving User Classification via Communication-Based Name and Location Clustering on Twitter. In: North American Chapter of the Association for Computational Linguistics (NAACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{bergsma:2013b, title = {Broadly Improving User Classification via Communication-Based Name and Location Clustering on Twitter}, author = {Shane Bergsma and Mark Dredze and Benjamin Van Durme and Theresa Wilson and David Yarowsky}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, abstract = {Hidden properties of social media users, such as their ethnicity, gender, and location, are often reflected in their observed attributes, such as their first and last names. Furthermore, users who communicate with each other often have similar hidden properties. We propose an algorithm that exploits these insights to cluster the observed attributes of hundreds of millions of Twitter users. Attributes such as user names are grouped together if users with those names communicate with other similar users. We separately cluster millions of unique first names, last names, and userprovided locations. The efficacy of these clusters is then evaluated on a diverse set of classification tasks that predict hidden users properties such as ethnicity, geographic location, gender, language, and race, using only profile names and locations when appropriate. Our readily-replicable approach and publicly released clusters are shown to be remarkably effective and versatile, substantially outperforming state-of-the-art approaches and human accuracy on each of the tasks studied.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Hidden properties of social media users, such as their ethnicity, gender, and location, are often reflected in their observed attributes, such as their first and last names. Furthermore, users who communicate with each other often have similar hidden properties. We propose an algorithm that exploits these insights to cluster the observed attributes of hundreds of millions of Twitter users. Attributes such as user names are grouped together if users with those names communicate with other similar users. We separately cluster millions of unique first names, last names, and userprovided locations. The efficacy of these clusters is then evaluated on a diverse set of classification tasks that predict hidden users properties such as ethnicity, geographic location, gender, language, and race, using only profile names and locations when appropriate. Our readily-replicable approach and publicly released clusters are shown to be remarkably effective and versatile, substantially outperforming state-of-the-art approaches and human accuracy on each of the tasks studied. |
| 530. | Joshi, Mahesh; Dredze, Mark; Cohen, William W; Rose, Carolyn P: What's in a Domain? Multi-Domain Learning for Multi-Attribute Data. In: North American Chapter of the Association for Computational Linguistics (NAACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{joshi:2013b, title = {What's in a Domain? Multi-Domain Learning for Multi-Attribute Data}, author = {Mahesh Joshi and Mark Dredze and William W. Cohen and Carolyn P. Rose}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, abstract = {Multi-Domain learning assumes that a single metadata attribute is used in order to divide the data into so-called domains. However, real-world datasets often have multiple metadata attributes that can divide the data into domains. It is not always apparent which single attribute will lead to the best domains, and more than one attribute might impact classification. We propose extensions to two multi-domain learning techniques for our multi-attribute setting, enabling them to simultaneously learn from several metadata attributes. Experimentally, they outperform the multi-domain learning baseline, even when it selects the single "best" attribute.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Multi-Domain learning assumes that a single metadata attribute is used in order to divide the data into so-called domains. However, real-world datasets often have multiple metadata attributes that can divide the data into domains. It is not always apparent which single attribute will lead to the best domains, and more than one attribute might impact classification. We propose extensions to two multi-domain learning techniques for our multi-attribute setting, enabling them to simultaneously learn from several metadata attributes. Experimentally, they outperform the multi-domain learning baseline, even when it selects the single "best" attribute. |
| 531. | Lamb, Alex; Paul, Michael J; Dredze, Mark: Separating Fact from Fear: Tracking Flu Infections on Twitter. In: North American Chapter of the Association for Computational Linguistics (NAACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{lamb:2013b, title = {Separating Fact from Fear: Tracking Flu Infections on Twitter}, author = {Alex Lamb and Michael J. Paul and Mark Dredze}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, abstract = {Twitter has been shown to be a fast and reliable method for disease surveillance of common illnesses like influenza. However, previous work has relied on simple content analysis, which conflates flu tweets that report infection with those that express concerned awareness of the flu. By discriminating these categories, as well as tweets about the authors versus about others, we demonstrate significant improvements on influenza surveillance using Twitter.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Twitter has been shown to be a fast and reliable method for disease surveillance of common illnesses like influenza. However, previous work has relied on simple content analysis, which conflates flu tweets that report infection with those that express concerned awareness of the flu. By discriminating these categories, as well as tweets about the authors versus about others, we demonstrate significant improvements on influenza surveillance using Twitter. |
| 532. | Paul, Michael; Dredze, Mark: Drug Extraction from the Web: Summarizing Drug Experiences with Multi-Dimensional Topic Models. In: North American Chapter of the Association for Computational Linguistics (NAACL), 2013. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Paul:2013b, title = {Drug Extraction from the Web: Summarizing Drug Experiences with Multi-Dimensional Topic Models}, author = {Michael Paul and Mark Dredze}, year = {2013}, date = {2013-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, abstract = {Multi-dimensional latent text models, such as factorial LDA (f-LDA), capture multiple factors of corpora, creating structured output for researchers to better understand the contents of a corpus. We consider such models for clinical research of new recreational drugs and trends, an important application for mining current information for healthcare workers. We use a "three-dimensional" f-LDA variant to jointly model combinations of drug (marijuana, salvia, etc.), aspect (effects, chemistry, etc.) and route of administration (smoking, oral, etc.) Since a purely unsupervised topic model is unlikely to discover these specific factors of interest, we develop a novel method of incorporating prior knowledge by leveraging user generated tags as priors in our model. We demonstrate that this model can be used as an exploratory tool for learning about these drugs from the Web by applying it to the task of extractive summarization. In addition to providing useful output for this important public health task, our prior-enriched model provides a framework for the application of f-LDA to other tasks}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Multi-dimensional latent text models, such as factorial LDA (f-LDA), capture multiple factors of corpora, creating structured output for researchers to better understand the contents of a corpus. We consider such models for clinical research of new recreational drugs and trends, an important application for mining current information for healthcare workers. We use a "three-dimensional" f-LDA variant to jointly model combinations of drug (marijuana, salvia, etc.), aspect (effects, chemistry, etc.) and route of administration (smoking, oral, etc.) Since a purely unsupervised topic model is unlikely to discover these specific factors of interest, we develop a novel method of incorporating prior knowledge by leveraging user generated tags as priors in our model. We demonstrate that this model can be used as an exploratory tool for learning about these drugs from the Web by applying it to the task of extractive summarization. In addition to providing useful output for this important public health task, our prior-enriched model provides a framework for the application of f-LDA to other tasks |
| 533. | Crammer, Koby; Kulesza, Alex; Dredze, Mark: Adaptive Regularization of Weight Vectors. In: Machine Learning, 91 (2), pp. 155-187, 2013. (Type: Journal Article | Abstract | Links | BibTeX) @article{Crammer:2013fkb, title = {Adaptive Regularization of Weight Vectors}, author = {Koby Crammer and Alex Kulesza and Mark Dredze}, doi = {10.1007/s10994-013-5327-x}, year = {2013}, date = {2013-01-01}, journal = {Machine Learning}, volume = {91}, number = {2}, pages = {155-187}, abstract = {We present AROW, an online learning algorithm for binary and multiclass problems that combines large margin training, confidence weighting, and the capacity to handle non-separable data. AROW performs adaptive regularization of the prediction function upon seeing each new instance, allowing it to perform especially well in the presence of label noise. We derive mistake bounds for the binary and multiclass settings that are similar in form to the second order perceptron bound. Our bounds do not assume separability. We also relate our algorithm to recent confidence-weighted online learning techniques. Empirical evaluations show that AROW achieves state-of-the-art performance on a wide range of binary and multiclass tasks, as well as robustness in the face of non-separable data.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We present AROW, an online learning algorithm for binary and multiclass problems that combines large margin training, confidence weighting, and the capacity to handle non-separable data. AROW performs adaptive regularization of the prediction function upon seeing each new instance, allowing it to perform especially well in the presence of label noise. We derive mistake bounds for the binary and multiclass settings that are similar in form to the second order perceptron bound. Our bounds do not assume separability. We also relate our algorithm to recent confidence-weighted online learning techniques. Empirical evaluations show that AROW achieves state-of-the-art performance on a wide range of binary and multiclass tasks, as well as robustness in the face of non-separable data. |
| 534. | Dredze, Mark H; Schilit, William N: Facet suggestion for search query augmentation. 2013, (US Patent 8,433,705). (Type: Patent | BibTeX) @patent{dredze2013facet, title = {Facet suggestion for search query augmentation}, author = {Mark H Dredze and William N Schilit}, year = {2013}, date = {2013-00-01}, publisher = {Google Patents}, note = {US Patent 8,433,705}, keywords = {}, pubstate = {published}, tppubtype = {patent} } |
| 535. | Dredze, Mark H; Schilit, William N: Facet suggestion for search query augmentation. 2013, (US Patent 8,433,705). (Type: Patent | BibTeX) @patent{dredze2013facetb, title = {Facet suggestion for search query augmentation}, author = {Mark H Dredze and William N Schilit}, year = {2013}, date = {2013-00-01}, publisher = {Google Patents}, note = {US Patent 8,433,705}, keywords = {}, pubstate = {published}, tppubtype = {patent} } |
| 536. | Dredze, Mark H; Schilit, William N: Facet suggestion for search query augmentation. 2013, (US Patent 8,433,705). (Type: Patent | BibTeX) @patent{dredze2013facetc, title = {Facet suggestion for search query augmentation}, author = {Mark H Dredze and William N Schilit}, year = {2013}, date = {2013-00-01}, publisher = {Google Patents}, note = {US Patent 8,433,705}, keywords = {}, pubstate = {published}, tppubtype = {patent} } |
| 537. | Paul, Michael J; Dredze, Mark: Experimenting with Drugs (and Topic Models): Multi-Dimensional Exploration of Recreational Drug Discussions. In: AAAI Fall Symposium on Information Retrieval and Knowledge Discovery in Biomedical Text, 2012. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Paul:2012fkc, title = {Experimenting with Drugs (and Topic Models): Multi-Dimensional Exploration of Recreational Drug Discussions}, author = {Michael J Paul and Mark Dredze}, year = {2012}, date = {2012-01-01}, booktitle = {AAAI Fall Symposium on Information Retrieval and Knowledge Discovery in Biomedical Text}, abstract = {Clinical research of new recreational drugs and trends requires mining current information from non-traditional text sources. In this work we support such research through the use of a multi-dimensional latent text model -- factorial LDA -- that captures orthogonal factors of corpora, creating structured output for researchers to better understand the contents of a corpus. Since a purely unsupervised model is unlikely to discover specific factors of interest to clinical researchers, we modify the structure of factorial LDA to incorporate prior knowledge, including the use of of observed variables, informative priors and background components. The resulting model learns factors that correspond to drug type, delivery method (smoking, injection, etc.), and aspect (chemistry, culture, effects, health, usage). We demonstrate that the improved model yields better quantitative and more interpretable results.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Clinical research of new recreational drugs and trends requires mining current information from non-traditional text sources. In this work we support such research through the use of a multi-dimensional latent text model -- factorial LDA -- that captures orthogonal factors of corpora, creating structured output for researchers to better understand the contents of a corpus. Since a purely unsupervised model is unlikely to discover specific factors of interest to clinical researchers, we modify the structure of factorial LDA to incorporate prior knowledge, including the use of of observed variables, informative priors and background components. The resulting model learns factors that correspond to drug type, delivery method (smoking, injection, etc.), and aspect (chemistry, culture, effects, health, usage). We demonstrate that the improved model yields better quantitative and more interpretable results. |
| 538. | Karakos, Damianos; Roark, Brian; Shafran, Izhak; Sagae, Kenji; Lehr, Maider; Prud'hommeaux, Emily; Xu, Puyang; Glenn, Nathan; Khudanpur, Sanjeev; Saraclar, Murat; Bikel, Dan; Dredze, Mark; Callison-Burch, Chris; Cao, Yuan; Hall, Keith; Hasler, Eva; Koehn, Philip; Lopez, Adam; Post, Matt; Riley, Darcey: Deriving conversation-based features from unlabeled speech for discriminative language modeling. In: International Speech Communication Association (INTERSPEECH), 2012. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Karakos:2012fkc, title = {Deriving conversation-based features from unlabeled speech for discriminative language modeling}, author = {Damianos Karakos and Brian Roark and Izhak Shafran and Kenji Sagae and Maider Lehr and Emily Prud'hommeaux and Puyang Xu and Nathan Glenn and Sanjeev Khudanpur and Murat Saraclar and Dan Bikel and Mark Dredze and Chris Callison-Burch and Yuan Cao and Keith Hall and Eva Hasler and Philip Koehn and Adam Lopez and Matt Post and Darcey Riley}, year = {2012}, date = {2012-01-01}, booktitle = {International Speech Communication Association (INTERSPEECH)}, abstract = {The perceptron algorithm was used in [1] to estimate discriminative language models which correct errors in the output of ASR systems. In its simplest version, the algorithm simply increases the weight of n-gram features which appear in the correct (oracle) hypothesis and decreases the weight of n-gram features which appear in the 1-best hypothesis. In this paper, we show that the perceptron algorithm can be successfully used in a semi-supervised learning (SSL) framework, where limited amounts of labeled data are available. Our framework has some similarities to graph-based label propagation [2] in the sense that a graph is built based on proximity of unlabeled conversations, and then it is used to propagate confidences (in the form of features) to the labeled data, based on which perceptron trains a discriminative model. The novelty of our approach lies in the fact that the confidence "flows" from the unlabeled data to the labeled data, and not vice-versa, as is done traditionally in SSL. Experiments conducted at the 2011 CLSP Summer Workshop on the conversational telephone speech corpora Dev04f and Eval04f demonstrate the effectiveness of the proposed approach.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The perceptron algorithm was used in [1] to estimate discriminative language models which correct errors in the output of ASR systems. In its simplest version, the algorithm simply increases the weight of n-gram features which appear in the correct (oracle) hypothesis and decreases the weight of n-gram features which appear in the 1-best hypothesis. In this paper, we show that the perceptron algorithm can be successfully used in a semi-supervised learning (SSL) framework, where limited amounts of labeled data are available. Our framework has some similarities to graph-based label propagation [2] in the sense that a graph is built based on proximity of unlabeled conversations, and then it is used to propagate confidences (in the form of features) to the labeled data, based on which perceptron trains a discriminative model. The novelty of our approach lies in the fact that the confidence "flows" from the unlabeled data to the labeled data, and not vice-versa, as is done traditionally in SSL. Experiments conducted at the 2011 CLSP Summer Workshop on the conversational telephone speech corpora Dev04f and Eval04f demonstrate the effectiveness of the proposed approach. |
| 539. | Rastrow, Ariya; Dredze, Mark; Khudanpur, Sanjeev: Efficient Structured Language Modeling for Speech Recognition. In: International Speech Communication Association (INTERSPEECH), 2012. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Rastrow:2012g, title = {Efficient Structured Language Modeling for Speech Recognition}, author = {Ariya Rastrow and Mark Dredze and Sanjeev Khudanpur}, year = {2012}, date = {2012-01-01}, booktitle = {International Speech Communication Association (INTERSPEECH)}, abstract = {The structured language model (SLM) of [1] was one of the first to successfully integrate syntactic structure into language models. We extend the SLM framework in two new directions. First, we propose a new syntactic hierarchical interpolation that improves over previous approaches. Second, we develop a general information-theoretic algorithm for pruning the underlying Jelinek-Mercer interpolated LM used in [1], which substantially reduces the size of the LM, enabling us to train on large data. When combined with hill-climbing [2] the SLM is an accurate model, space-efficient and fast for rescoring large speech lattices. Experimental results on broadcast news demonstrate that the SLM outperforms a large 4-gram LM.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The structured language model (SLM) of [1] was one of the first to successfully integrate syntactic structure into language models. We extend the SLM framework in two new directions. First, we propose a new syntactic hierarchical interpolation that improves over previous approaches. Second, we develop a general information-theoretic algorithm for pruning the underlying Jelinek-Mercer interpolated LM used in [1], which substantially reduces the size of the LM, enabling us to train on large data. When combined with hill-climbing [2] the SLM is an accurate model, space-efficient and fast for rescoring large speech lattices. Experimental results on broadcast news demonstrate that the SLM outperforms a large 4-gram LM. |
| 540. | Green, Spence; Andrews, Nicholas; Gormley, Matthew; Dredze, Mark; Manning, Christopher D: Entity Clustering Across Languages. In: North American Chapter of the Association for Computational Linguistics (NAACL), pp. 60-69, 2012. (Type: Inproceedings | Abstract | BibTeX) @inproceedings{Green:2012uqd, title = {Entity Clustering Across Languages}, author = {Spence Green and Nicholas Andrews and Matthew Gormley and Mark Dredze and Christopher D Manning}, year = {2012}, date = {2012-01-01}, booktitle = {North American Chapter of the Association for Computational Linguistics (NAACL)}, pages = {60-69}, abstract = {Standard entity clustering systems commonly rely on mention (string) matching, syntactic features, and linguistic resources like English WordNet. When co-referent text mentions appear in different languages, these techniques cannot be easily applied. Consequently, we develop new methods for clustering text mentions across documents and languages simultaneously, producing cross-lingual entity clusters. Our approach extends standard clustering algorithms with cross-lingual mention and context similarity measures. Crucially, we do not assume a pre-existing entity list (knowledge base), so entity characteristics are unknown. On an Arabic-English corpus that contains seven different text genres, our best model yields a 24.3% F1 gain over the baseline.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Standard entity clustering systems commonly rely on mention (string) matching, syntactic features, and linguistic resources like English WordNet. When co-referent text mentions appear in different languages, these techniques cannot be easily applied. Consequently, we develop new methods for clustering text mentions across documents and languages simultaneously, producing cross-lingual entity clusters. Our approach extends standard clustering algorithms with cross-lingual mention and context similarity measures. Crucially, we do not assume a pre-existing entity list (knowledge base), so entity characteristics are unknown. On an Arabic-English corpus that contains seven different text genres, our best model yields a 24.3% F1 gain over the baseline. |
Eisner, Jason Indirect STV Election: A Voting System for South Africa Miscellaneous 1991. Many modern students of democracy favor proportional representation through the Single Transferable Vote (STV). In countries with high illiteracy, however, this system may be unworkable. This paper proposes a practical modification of STV. In the modified system, each citizen votes for only one candidate. Voters need not specify their second, third, and fourth choices. Instead, each candidate specifies his or her second, third, and fourth choices. The modified system is no more difficult for voters than current proposals -- and it provides virtually all the benefits of STV, together with some new ones.}, Eisner, Jason Cognitive Science and the Search for Intelligence Miscellaneous 1991. Bayya, Bayya; Hermansky, Hynek Towards Feature-Based Speech Metric Inproceedings Proceedings of the IEEE International Conference on Acoustics, Albuquerque, NM, 1990. Eisner, Jason Dynamical-Systems Behavior in Recurrent and Non-Recurrent Connectionist Nets Miscellaneous 1990. Hermansky, Hynek; Broad, D The Effective Second Formant F2' and the Vocal Tract Front Cavity Inproceedings Proceedings of the IEEE International Conference on Acoustics, Glasgow, Scotland, 1989. Boahen, KA; Andreou, Andreas; Pouliquen, PO; Pavasovic, A Architectures for associative memories using current-mode analog MOS circuits Journal Article pp. 175–193, 1989. Boahen, KA; Pouliquen, PO; Andreou, Andreas; Jenkins, RE A heteroassociative memory using current-mode MOS analog VLSI circuits Journal Article pp. 747–755, 1989. Hermansky, Hynek; Junqua, J Optimization of Perceptually Based ASR Front End Inproceedings Proceedings of the IEEE International Conference on Acoustics, York, NY, 1988. Khudanpur, Sanjeev Design and Implementation of an Ethernet Communications Controller for IBM Compatible Personal Computers Technical Report Bombay, India, 1988. Hermansky, Hynek Automatic Speech Recognition and Human Auditory Perception Inproceedings Proceedings of the European Conference on Speech Communication and Technology, Edinburgh, Scotland, 1987. Hermansky, Hynek Role of Relative Positions of Formant and Harmonic Peaks in Perception of Vowel-Like Stimuli Technical Report Barbara, CA, 1987. Hermansky, Hynek An Efficient Speaker-Independent Automatic Speech Recognition By Simulation of Some Properties of Human Auditory Processing Inproceedings Proceedings of the IEEE International Conference on Acoustics, Dallas, TX, 1987. Javkin, H; Hermansky, Hynek; Wakita, H Interaction Between Formant and Harmonic Peaks in Vowel Perception Inproceedings Proceedings of Eleventh International Congress of Phonetic Sciences, Tallin, Estonia, 1987. Hermansky, Hynek; Tsuga, K; Makino, S; Wakita, H Perceptually Based Processing in Automatic Speech Recognition Inproceedings Proceedings of the IEEE International Conference on Acoustics, Tokyo, Japan, 1986. Hermansky, Hynek; Hanson, B; Wakita, H Perceptually Based Linear Predictive Analysis of Speech Inproceedings Proceedings of the IEEE International Conference on Acoustics, Tampa, FL, 1985. Hermansky, Hynek; Hanson, B; Wakita, H; Fujisaki, H Linear Predictive Modeling of Speech in Modified Spectral Domains Inproceedings Proceedings of International Conference on Digital Processg of Signals Communications, Loughbrorough, England, 1985. Hermansky, Hynek; Fujisaki, H; Sato, Y Spectral Envelope Sampling and Interpolation in Linear Predictive Analysis of Speech Inproceedings Proceedings of the IEEE International Conference on Acoustics, Diego, CA, 1984. Hermansky, Hynek; Fujisaki, H; Sato, Y Analysis and Synthesis of Speech Based on Spectral Transform Linear Predictive Method Inproceedings Proceedings of the IEEE International Conference on Acoustics, Boston, MA, 1983. Hermansky, Hynek; Fujisaki, H; Sato, Y Spectral Transform Linear Predictive Analysis for Analysis-Synthesis of Speech Inproceedings Proceedings of 11e Congress International d'Acoustique, Paris, France, 1983. Hermansky, Hynek; Fujisaki, H; Sato, Y Speech Analysis-Synthesis System Based on STLP Method Inproceedings Proceedings of the Sprg Meetg of the Acoustical Society of Japan, Japan, Tokyo, 1983. Hermansky, Hynek; Fujisaki, H; Sato, Y On Analysis of Real Speech Based on STLP Method Inproceedings Proceedings of the Sprg Meetg of the Acoustical Society of Japan, Nagaoka, Japan, 1982. Hermansky, Hynek; Fujisaki, H; Sato, Y Spectral Envelope Sampling in Linear Predictive Analysis of Speech Inproceedings Proceedings of the Sprg Meetg of the Acoustical Society of Japan, Tokyo, Japan, 1982. Hermansky, Hynek; Fujisaki, H Spectral Transforms in Linear Prediction Inproceedings Proceedings of the Annual Meetg of the Institute of Electronics and Communication Engeers of Japan, Tokyo, Japan, 1982. Hermansky, Hynek Improved Linear Predictive Analysis of Speech Based on Spectral Processing Inproceedings Dr. Eng. Thesis, Tokyo, Japan, 1982. Hermansky, Hynek; Fujisaki, H; Sato, Y Sampling and Interpolation of Spectral Envelopes in Linear Predictive Analysis of Speech Inproceedings Transactions of the Committee on Speech Research, pp. 97-104, 1982. Hermansky, Hynek; Fujisaki, H Spectral Transforms in Linear Predictive Analysis Inproceedings Proceedings of the Fall Meetg of the Acoustical Society of Japan, Kagoshima, Japan, 1981. Hermansky, Hynek; Fujisaki, H LPC Methods in the Short Time Analysis of Voiced Speech Inproceedings Proceedings of the Sprg Meetg of the Acoustical Society of Japan, Japan, Tokyo, 1981. Hermansky, Hynek; Fujisaki, H The Effect of Spectral Transforms in Linear Predictive Analysis of Speech Journal Article pp. 365-372, 1981. Hermansky, Hynek; Fujisaki, H Acoustic Characteristics of Czech Vowels, Research on Human Information Processing Inproceedings Proceedings of the Fall Meetg of the Acoustical Society of Japan, Shimizu, Japan, 1980. Rush, A M; Sontag, D; Collins, M; Jaakkola, T; Huang, Z; Harper, M; Petrov, S; Cheung, J C K; Penn, G; Martins, A; others others, Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing$$ Journal Article 0000. Sakaguchi, Keisuke ; Post, Matt ; Van Durme, Benjamin Efficient Elicitation of Annotations for Human Evaluation of Machine Translation Inproceedings Proceedings of the Ninth Workshop on Statistical Machine Translation, pp. 1–11, Association for Computational Linguistics, Baltimore, Maryland, USA, 0000. Manohar, Vimal; Srinivas, C B; Umesh, S Acoustic modeling using transform-based phone-cluster adaptive training Inproceedings Automatic Speech Recognition and Understanding (ASRU), 2013 IEEE Workshop on, pp. 49-54, 0000. Blaschko, Matthew; Girshick, Ross B; Kannala, Juho; Kokkinos, Iasonas; Mahendran, Siddarth; Maji, Subhransu; Mohamed, Sammy; Rahtu, Esa; Saphra, Naomi ; Simonyan, Karen; others others, Towards a detailed understanding of objects and scenes in natural images Journal Article 0000. Rudinger, Rachel ; Van Durme, Benjamin Is the Stanford Dependency Representation Semantic? Inproceedings Association for Computational Linguistics (ACL), Workshop on EVENTS, 0000. Saphra, Naomi; Lopez, Adam AMRICA: an AMR Inspector for Cross-language Alignments Inproceedings Proc. of NAACL, 0000. Khudanpur, Sanjeev; Povey, Daniel; Peddinti, Vijayaditya; Seltzer, Michael L A study on data augmentation of reverberant speech for robust speech recognition Conference 0000. Snyder, David; Ghahremani, Pegah; Povey, Daniel; Garcia-Romero, Daniel; Carmiel, Yishay; Khudanpur, Sanjeev DEEP NEURAL NETWORK-BASED SPEAKER EMBEDDINGS FOR END-TO-END SPEAKER VERIFICATION Journal Article 0000. Peddinti, Vimal Manohar Vijayaditya; Wang, Yiming; Povey, Daniel; Khudanpur, Sanjeev Far-field ASR without parallel data Journal Article 0000. Hadian, Hossein; Sameti, Hossein; Povey, Daniel; Khudanpur, Sanjeev TOWARDS DISCRIMINATIVELY-TRAINED HMM-BASED END-TO-END MODELS FOR AUTOMATIC SPEECH RECOGNITION Journal Article 0000. Ghahremani, Pegah; Manohar, Vimal; Hadian, Hossein; Povey, Daniel; Khudanpur, Sanjeev INVESTIGATION OF TRANSFER LEARNING FOR ASR USING LF-MMI TRAINED NEURAL NETWORKS Journal Article 0000. Hori, Takaaki; Watanabe, Shinji; Zhang, Yu; Chan, William Advances in Joint CTC-Attention based End-to-End Speech Recognition with a Deep CNN Encoder and RNN-LM Inproceedings Proc. Interspeech 2017, pp. 949–953 year=2017, 0000. Wang, Dingquan; Eisner, Jason Surface Statistics of an Unknown Language Indicate How
to Parse It Journal Article Transactions of the Association for Computational
Linguistics (TACL), 0000, ISSN: 2307-387X. Paul, Michael J; Chisolm, Margaret S; Johnson, Matthew W; Vandrey, Ryan G; Dredze, Mark Assessing the validity of online drug forums as a source for estimating demographic and temporal trends in drug use Journal Article Journal of Addiction Medicine, 10 (5), pp. 324–330, 0000. Dredze, Mark; Schilit, Bill Facet suggestion for search query augmentation Patent 0000, (US Patent 8,433,705). Knowles, Rebecca; Koehn, Philipp Context and Copying in Neural Machine Translation Inproceedings Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pp. 3034–3041, Association for Computational Linguistics, Brussels, Belgium, 0000.
1991
title = {Indirect STV Election: A Voting System for South Africa},
author = {Jason Eisner},
year = {1991},
date = {1991-01-01},
abstract = {"Winner take all" electoral systems are not fully representative. Unfortunately, the ANC's proposed system of proportional representation is not much better. Because it ensconces party politics, it is only slightly more representative, and poses a serious threat to accountability.
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
title = {Cognitive Science and the Search for Intelligence},
author = {Jason Eisner},
year = {1991},
date = {1991-01-01},
abstract = {This talk for a general audience gives a sketch of what the field of cognitive science is about. In its latter half, it turns to the philosophical question of defining intelligence, and proposes a non-operational alternative to the Turing Test.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
1990
title = {Towards Feature-Based Speech Metric},
author = {Bayya Bayya and Hynek Hermansky},
year = {1990},
date = {1990-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Albuquerque, NM},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Dynamical-Systems Behavior in Recurrent and Non-Recurrent Connectionist Nets},
author = {Jason Eisner},
year = {1990},
date = {1990-01-01},
abstract = {A broad approach is developed for training dynamical behaviors in connectionist networks. General recurrent networks are powerful computational devices, necessary for difficult tasks like constraint satisfaction and temporal processing. These tasks are discussed here in some detail. From both theoretical and empirical considerations, it is concluded that such tasks are best addressed by recurrent networks that operate continuously in time -- and further, that effective learning rules for these continuous-time networks must be able to prescribe their textitdynamical properties. A general class of such learning rules is derived and tested on simple problems. Where existing learning algorithms for recurrent and non-recurrent networks only attempt to train a network's textitposition in activation space, the models presented here can also explicitly and successfully prescribe the nature of its textitmovement through activation space.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
1989
title = {The Effective Second Formant F2' and the Vocal Tract Front Cavity},
author = {Hynek Hermansky and D. Broad},
year = {1989},
date = {1989-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Glasgow, Scotland},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Architectures for associative memories using current-mode analog MOS circuits},
author = {KA Boahen and Andreas Andreou and PO Pouliquen and A Pavasovic},
year = {1989},
date = {1989-01-01},
pages = {175--193},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {A heteroassociative memory using current-mode MOS analog VLSI circuits},
author = {KA Boahen and PO Pouliquen and Andreas Andreou and RE Jenkins},
year = {1989},
date = {1989-01-01},
pages = {747--755},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
1988
title = {Optimization of Perceptually Based ASR Front End},
author = {Hynek Hermansky and J. Junqua},
year = {1988},
date = {1988-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {York, NY},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Design and Implementation of an Ethernet Communications Controller for IBM Compatible Personal Computers},
author = {Sanjeev Khudanpur},
year = {1988},
date = {1988-01-01},
publisher = {Bachelor of Technology Project, Department of Electrical Engineering},
address = {Bombay, India},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
1987
title = {Automatic Speech Recognition and Human Auditory Perception},
author = {Hynek Hermansky},
year = {1987},
date = {1987-01-01},
booktitle = {Proceedings of the European Conference on Speech Communication and Technology},
address = {Edinburgh, Scotland},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Role of Relative Positions of Formant and Harmonic Peaks in Perception of Vowel-Like Stimuli},
author = {Hynek Hermansky},
year = {1987},
date = {1987-01-01},
booktitle = {STL Technical Reports},
address = {Barbara, CA},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
title = {An Efficient Speaker-Independent Automatic Speech Recognition By Simulation of Some Properties of Human Auditory Processing},
author = {Hynek Hermansky},
year = {1987},
date = {1987-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Dallas, TX},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Interaction Between Formant and Harmonic Peaks in Vowel Perception},
author = {H. Javkin and Hynek Hermansky and H. Wakita},
year = {1987},
date = {1987-01-01},
booktitle = {Proceedings of Eleventh International Congress of Phonetic Sciences},
address = {Tallin, Estonia},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1986
title = {Perceptually Based Processing in Automatic Speech Recognition},
author = {Hynek Hermansky and K. Tsuga and S. Makino and H. Wakita},
year = {1986},
date = {1986-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Tokyo, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1985
title = {Perceptually Based Linear Predictive Analysis of Speech},
author = {Hynek Hermansky and B. Hanson and H. Wakita},
year = {1985},
date = {1985-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Tampa, FL},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Linear Predictive Modeling of Speech in Modified Spectral Domains},
author = {Hynek Hermansky and B. Hanson and H. Wakita and H. Fujisaki},
year = {1985},
date = {1985-01-01},
booktitle = {Proceedings of International Conference on Digital Processg of Signals Communications},
address = {Loughbrorough, England},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1984
title = {Spectral Envelope Sampling and Interpolation in Linear Predictive Analysis of Speech},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1984},
date = {1984-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Diego, CA},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1983
title = {Analysis and Synthesis of Speech Based on Spectral Transform Linear Predictive Method},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1983},
date = {1983-01-01},
booktitle = {Proceedings of the IEEE International Conference on Acoustics},
address = {Boston, MA},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Spectral Transform Linear Predictive Analysis for Analysis-Synthesis of Speech},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1983},
date = {1983-01-01},
booktitle = {Proceedings of 11e Congress International d'Acoustique},
address = {Paris, France},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Speech Analysis-Synthesis System Based on STLP Method},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1983},
date = {1983-01-01},
booktitle = {Proceedings of the Sprg Meetg of the Acoustical Society of Japan},
address = {Japan, Tokyo},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1982
title = {On Analysis of Real Speech Based on STLP Method},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1982},
date = {1982-01-01},
booktitle = {Proceedings of the Sprg Meetg of the Acoustical Society of Japan},
address = {Nagaoka, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Spectral Envelope Sampling in Linear Predictive Analysis of Speech},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1982},
date = {1982-01-01},
booktitle = {Proceedings of the Sprg Meetg of the Acoustical Society of Japan},
address = {Tokyo, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Spectral Transforms in Linear Prediction},
author = {Hynek Hermansky and H. Fujisaki},
year = {1982},
date = {1982-01-01},
booktitle = {Proceedings of the Annual Meetg of the Institute of Electronics and Communication Engeers of Japan},
address = {Tokyo, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Improved Linear Predictive Analysis of Speech Based on Spectral Processing},
author = {Hynek Hermansky},
year = {1982},
date = {1982-01-01},
booktitle = {Dr. Eng. Thesis},
address = {Tokyo, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Sampling and Interpolation of Spectral Envelopes in Linear Predictive Analysis of Speech},
author = {Hynek Hermansky and H. Fujisaki and Y. Sato},
year = {1982},
date = {1982-01-01},
booktitle = {Transactions of the Committee on Speech Research},
pages = {97-104},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
1981
title = {Spectral Transforms in Linear Predictive Analysis},
author = {Hynek Hermansky and H. Fujisaki},
year = {1981},
date = {1981-01-01},
booktitle = {Proceedings of the Fall Meetg of the Acoustical Society of Japan},
address = {Kagoshima, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {LPC Methods in the Short Time Analysis of Voiced Speech},
author = {Hynek Hermansky and H. Fujisaki},
year = {1981},
date = {1981-01-01},
booktitle = {Proceedings of the Sprg Meetg of the Acoustical Society of Japan},
address = {Japan, Tokyo},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {The Effect of Spectral Transforms in Linear Predictive Analysis of Speech},
author = {Hynek Hermansky and H. Fujisaki},
year = {1981},
date = {1981-01-01},
booktitle = {Transactions of the Committee on Speech Research},
pages = {365-372},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
1980
title = {Acoustic Characteristics of Czech Vowels, Research on Human Information Processing},
author = {Hynek Hermansky and H. Fujisaki},
year = {1980},
date = {1980-01-01},
booktitle = {Proceedings of the Fall Meetg of the Acoustical Society of Japan},
address = {Shimizu, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
0000
title = {Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing$$},
author = {A.M. Rush and D. Sontag and M. Collins and T. Jaakkola and Z. Huang and M. Harper and S. Petrov and J.C.K. Cheung and G. Penn and A. Martins and others others},
publisher = {Association for Computational Linguistics$$},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Efficient Elicitation of Annotations for Human Evaluation of Machine Translation},
author = {Sakaguchi, Keisuke and Post, Matt and Van Durme, Benjamin},
url = {http://www.aclweb.org/anthology/W14-3301},
booktitle = {Proceedings of the Ninth Workshop on Statistical Machine Translation},
pages = {1--11},
publisher = {Association for Computational Linguistics},
address = {Baltimore, Maryland, USA},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Acoustic modeling using transform-based phone-cluster adaptive training},
author = {Vimal Manohar and C.B. Srinivas and S. Umesh},
booktitle = {Automatic Speech Recognition and Understanding (ASRU), 2013 IEEE Workshop on},
pages = {49-54},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Towards a detailed understanding of objects and scenes in natural images},
author = {Matthew Blaschko and Ross B. Girshick and Juho Kannala and Iasonas Kokkinos and Siddarth Mahendran and Subhransu Maji and Sammy Mohamed and Esa Rahtu and Saphra, Naomi and Karen Simonyan and others others},
url = {http://hal.archives-ouvertes.fr/hal-00776027/},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Is the Stanford Dependency Representation Semantic?},
author = {Rudinger, Rachel and Van Durme, Benjamin},
url = {http://www.ling.uni-potsdam.de/~koller/aclpub/W14-29/cdrom/W14-29-2014.pdf#page=64},
booktitle = {Association for Computational Linguistics (ACL), Workshop on EVENTS},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {AMRICA: an AMR Inspector for Cross-language Alignments},
author = {Naomi Saphra and Lopez, Adam},
booktitle = {Proc. of NAACL},
abstract = {Abstract Meaning Representation (AMR), an annotation scheme for natural language semantics, has drawn attention for its simplicity and representational power. Because AMR annotations are not designed for human readability, we present AMRICA, a visual aid for exploration of AMR annotations. AMRICA can visualize an AMR or the difference between two AMRs to help users diagnose interannotator disagreement or errors from an AMR parser. AMRICA can also automatically align and visualize the AMRs of a sentence and its translation in a parallel text. We believe AMRICA will simplify and streamline exploratory research on cross-lingual AMR corpora.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {A study on data augmentation of reverberant speech for robust speech recognition},
author = {Sanjeev Khudanpur and Daniel Povey and Vijayaditya Peddinti and Michael L. Seltzer },
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
title = {DEEP NEURAL NETWORK-BASED SPEAKER EMBEDDINGS FOR END-TO-END SPEAKER VERIFICATION},
author = {David Snyder and Pegah Ghahremani and Daniel Povey and Daniel Garcia-Romero and Yishay Carmiel and Sanjeev Khudanpur},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Far-field ASR without parallel data},
author = {Vimal Manohar Vijayaditya Peddinti and Yiming Wang and Daniel Povey and Sanjeev Khudanpur},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {TOWARDS DISCRIMINATIVELY-TRAINED HMM-BASED END-TO-END MODELS FOR AUTOMATIC SPEECH RECOGNITION},
author = {Hossein Hadian and Hossein Sameti and Daniel Povey and Sanjeev Khudanpur},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {INVESTIGATION OF TRANSFER LEARNING FOR ASR USING LF-MMI TRAINED NEURAL NETWORKS},
author = {Pegah Ghahremani and Vimal Manohar and Hossein Hadian and Daniel Povey and Sanjeev Khudanpur},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Advances in Joint CTC-Attention based End-to-End Speech Recognition with a Deep CNN Encoder and RNN-LM},
author = {Takaaki Hori and Shinji Watanabe and Yu Zhang and William Chan},
booktitle = {Proc. Interspeech 2017},
pages = {949--953 year=2017},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
title = {Surface Statistics of an Unknown Language Indicate How
to Parse It},
author = {Dingquan Wang and Jason Eisner},
url = {http://cs.jhu.edu/~jason/papers/#wang-eisner-2018-tacl},
issn = {2307-387X},
journal = {Transactions of the Association for Computational
Linguistics (TACL)},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Assessing the validity of online drug forums as a source for estimating demographic and temporal trends in drug use},
author = {Michael J Paul and Margaret S Chisolm and Matthew W Johnson and Ryan G Vandrey and Mark Dredze},
journal = {Journal of Addiction Medicine},
volume = {10},
number = {5},
pages = {324--330},
abstract = {Objectives: Addiction researchers have begun monitoring online forums to uncover self-reported details about use and effects of emerging drugs. The use of such online data sources has not been validated against data from large epidemiological surveys. This study aimed to characterize and compare the demographic and temporal trends associated with drug use as reported in online forums and in a large epidemiological survey. Methods: Data were collected from the website, drugs-forum.com, from January 2007 through August 2012 (143,416 messages posted by 8,087 members) and from the United States National Survey on Drug Use and Health (NSDUH) from 2007-2012. Measures of forum participation levels were compared with and validated against two measures from the NSDUH survey data: percentage of people using the drug in last 30 days and percentage using the drug more than 100 times in the past year. Results: For established drugs (e.g., cannabis), significant correlations were found across demographic groups between drugs-forum.com and the NSDUH survey data, while weaker, non-significant correlations were found with temporal trends. Emerging drugs (e.g., Salvia divinorum) were strongly associated with male users in the forum, in agreement with survey-derived data, and had temporal patterns that increased in synchrony with poison control reports. Conclusions: These results offer the first assessment of online drug forums as a valid source for estimating demographic and temporal trends in drug use. The analyses suggest that online forums are a reliable source for estimation of demographic associations and early identification of emerging drugs, but a less reliable source for measurement of long-term temporal trends.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
title = {Facet suggestion for search query augmentation},
author = {Mark Dredze and Bill Schilit},
publisher = {Google Patents},
note = {US Patent 8,433,705},
keywords = {},
pubstate = {published},
tppubtype = {patent}
}
title = {Context and Copying in Neural Machine Translation},
author = {Rebecca Knowles and Philipp Koehn},
url = {http://www.aclweb.org/anthology/D18-1339},
booktitle = {Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing},
pages = {3034--3041},
publisher = {Association for Computational Linguistics},
address = {Brussels, Belgium},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Upcoming Seminars
There are no upcoming events.
