Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Lyle Ungar

Lyle Ungar

University of Pennsylvania

Social Media Reveals Urban-Rural Differences in Stress across China


Nov 03, 2021
Jesse Cui, Tingdan Zhang, Kokil Jaidka, Dandan Pang, Garrick Sherman, Vinit Jakhetiya, Lyle Ungar, Sharath Chandra Guntuku

Figure 1 for Social Media Reveals Urban-Rural Differences in Stress across China
Figure 2 for Social Media Reveals Urban-Rural Differences in Stress across China
Figure 3 for Social Media Reveals Urban-Rural Differences in Stress across China
Figure 4 for Social Media Reveals Urban-Rural Differences in Stress across China
* Accepted at AAAI Conference on Web and Social Media (ICWSM) 2022 
  Access Paper or Ask Questions

Does BERT Learn as Humans Perceive? Understanding Linguistic Styles through Lexica


Sep 06, 2021
Shirley Anugrah Hayati, Dongyeop Kang, Lyle Ungar

Figure 1 for Does BERT Learn as Humans Perceive? Understanding Linguistic Styles through Lexica
Figure 2 for Does BERT Learn as Humans Perceive? Understanding Linguistic Styles through Lexica
Figure 3 for Does BERT Learn as Humans Perceive? Understanding Linguistic Styles through Lexica
Figure 4 for Does BERT Learn as Humans Perceive? Understanding Linguistic Styles through Lexica
* Accepted at EMNLP 2021 Main Conference 
  Access Paper or Ask Questions

Toward Micro-Dialect Identification in Diaglossic and Code-Switched Environments


Oct 10, 2020
Muhammad Abdul-Mageed, Chiyu Zhang, AbdelRahim Elmadany, Lyle Ungar

Figure 1 for Toward Micro-Dialect Identification in Diaglossic and Code-Switched Environments
Figure 2 for Toward Micro-Dialect Identification in Diaglossic and Code-Switched Environments
Figure 3 for Toward Micro-Dialect Identification in Diaglossic and Code-Switched Environments
Figure 4 for Toward Micro-Dialect Identification in Diaglossic and Code-Switched Environments
* Accepted in EMNLP 2020 
  Access Paper or Ask Questions

Generalized SHAP: Generating multiple types of explanations in machine learning


Jun 15, 2020
Dillon Bowen, Lyle Ungar

Figure 1 for Generalized SHAP: Generating multiple types of explanations in machine learning
Figure 2 for Generalized SHAP: Generating multiple types of explanations in machine learning
Figure 3 for Generalized SHAP: Generating multiple types of explanations in machine learning
Figure 4 for Generalized SHAP: Generating multiple types of explanations in machine learning
* 12 pages, 7 figures. Based on a submission to NeurIPS 2020. Dillon Bowen is credited with the original concept, code, data analysis, and initial paper draft. Lyle Ungar is credited with contributions to the draft and mathematical notation. Documentation can be found at https://dsbowen.github.io/gshap/ 
  Access Paper or Ask Questions

Learning Word Ratings for Empathy and Distress from Document-Level User Responses


Dec 02, 2019
João Sedoc, Sven Buechel, Yehonathan Nachmany, Anneke Buffone, Lyle Ungar

Figure 1 for Learning Word Ratings for Empathy and Distress from Document-Level User Responses
Figure 2 for Learning Word Ratings for Empathy and Distress from Document-Level User Responses
Figure 3 for Learning Word Ratings for Empathy and Distress from Document-Level User Responses
Figure 4 for Learning Word Ratings for Empathy and Distress from Document-Level User Responses

  Access Paper or Ask Questions

Correcting Sociodemographic Selection Biases for Accurate Population Prediction from Social Media


Nov 10, 2019
Salvatore Giorgi, Veronica Lynn, Sandra Matz, Lyle Ungar, H. Andrew Schwartz

Figure 1 for Correcting Sociodemographic Selection Biases for Accurate Population Prediction from Social Media
Figure 2 for Correcting Sociodemographic Selection Biases for Accurate Population Prediction from Social Media
Figure 3 for Correcting Sociodemographic Selection Biases for Accurate Population Prediction from Social Media
Figure 4 for Correcting Sociodemographic Selection Biases for Accurate Population Prediction from Social Media

  Access Paper or Ask Questions

Sentence-Level BERT and Multi-Task Learning of Age and Gender in Social Media


Nov 02, 2019
Muhammad Abdul-Mageed, Chiyu Zhang, Arun Rajendran, AbdelRahim Elmadany, Michael Przystupa, Lyle Ungar

Figure 1 for Sentence-Level BERT and Multi-Task Learning of Age and Gender in Social Media
Figure 2 for Sentence-Level BERT and Multi-Task Learning of Age and Gender in Social Media
Figure 3 for Sentence-Level BERT and Multi-Task Learning of Age and Gender in Social Media
Figure 4 for Sentence-Level BERT and Multi-Task Learning of Age and Gender in Social Media

  Access Paper or Ask Questions

DiaNet: BERT and Hierarchical Attention Multi-Task Learning of Fine-Grained Dialect


Oct 31, 2019
Muhammad Abdul-Mageed, Chiyu Zhang, AbdelRahim Elmadany, Arun Rajendran, Lyle Ungar

Figure 1 for DiaNet: BERT and Hierarchical Attention Multi-Task Learning of Fine-Grained Dialect
Figure 2 for DiaNet: BERT and Hierarchical Attention Multi-Task Learning of Fine-Grained Dialect
Figure 3 for DiaNet: BERT and Hierarchical Attention Multi-Task Learning of Fine-Grained Dialect
Figure 4 for DiaNet: BERT and Hierarchical Attention Multi-Task Learning of Fine-Grained Dialect

  Access Paper or Ask Questions

Conceptor Debiasing of Word Representations Evaluated on WEAT


Jun 14, 2019
Saket Karve, Lyle Ungar, João Sedoc

Figure 1 for Conceptor Debiasing of Word Representations Evaluated on WEAT
Figure 2 for Conceptor Debiasing of Word Representations Evaluated on WEAT
Figure 3 for Conceptor Debiasing of Word Representations Evaluated on WEAT
Figure 4 for Conceptor Debiasing of Word Representations Evaluated on WEAT

  Access Paper or Ask Questions

Continual Learning for Sentence Representations Using Conceptors


Apr 18, 2019
Tianlin Liu, Lyle Ungar, João Sedoc

Figure 1 for Continual Learning for Sentence Representations Using Conceptors
Figure 2 for Continual Learning for Sentence Representations Using Conceptors
* Accepted by NAACL-2019 
  Access Paper or Ask Questions

Unsupervised Post-processing of Word Vectors via Conceptor Negation


Dec 02, 2018
Tianlin Liu, Lyle Ungar, João Sedoc

Figure 1 for Unsupervised Post-processing of Word Vectors via Conceptor Negation
Figure 2 for Unsupervised Post-processing of Word Vectors via Conceptor Negation
Figure 3 for Unsupervised Post-processing of Word Vectors via Conceptor Negation
Figure 4 for Unsupervised Post-processing of Word Vectors via Conceptor Negation
* Accepted by AAAI-2019 
  Access Paper or Ask Questions

Understanding and Measuring Psychological Stress using Social Media


Nov 19, 2018
Sharath Chandra Guntuku, Anneke Buffone, Kokil Jaidka, Johannes Eichstaedt, Lyle Ungar

Figure 1 for Understanding and Measuring Psychological Stress using Social Media
Figure 2 for Understanding and Measuring Psychological Stress using Social Media
Figure 3 for Understanding and Measuring Psychological Stress using Social Media
Figure 4 for Understanding and Measuring Psychological Stress using Social Media
* Accepted for publication in the proceedings of ICWSM 2019 
  Access Paper or Ask Questions

Correcting the Common Discourse Bias in Linear Representation of Sentences using Conceptors


Nov 17, 2018
Tianlin Liu, João Sedoc, Lyle Ungar

Figure 1 for Correcting the Common Discourse Bias in Linear Representation of Sentences using Conceptors
* Accepted by the BioCreative/OHNLP workshop of ACM-BCB 2018 
  Access Paper or Ask Questions

Learning Neural Emotion Analysis from 100 Observations: The Surprising Effectiveness of Pre-Trained Word Representations


Oct 25, 2018
Sven Buechel, João Sedoc, H. Andrew Schwartz, Lyle Ungar

Figure 1 for Learning Neural Emotion Analysis from 100 Observations: The Surprising Effectiveness of Pre-Trained Word Representations
Figure 2 for Learning Neural Emotion Analysis from 100 Observations: The Surprising Effectiveness of Pre-Trained Word Representations
Figure 3 for Learning Neural Emotion Analysis from 100 Observations: The Surprising Effectiveness of Pre-Trained Word Representations
Figure 4 for Learning Neural Emotion Analysis from 100 Observations: The Surprising Effectiveness of Pre-Trained Word Representations

  Access Paper or Ask Questions

Modeling Empathy and Distress in Reaction to News Stories


Aug 30, 2018
Sven Buechel, Anneke Buffone, Barry Slaff, Lyle Ungar, João Sedoc

Figure 1 for Modeling Empathy and Distress in Reaction to News Stories
Figure 2 for Modeling Empathy and Distress in Reaction to News Stories
Figure 3 for Modeling Empathy and Distress in Reaction to News Stories
Figure 4 for Modeling Empathy and Distress in Reaction to News Stories
* To appear at EMNLP 2018 
  Access Paper or Ask Questions

Domain Aware Neural Dialog System


Aug 02, 2017
Sajal Choudhary, Prerna Srivastava, Lyle Ungar, João Sedoc

Figure 1 for Domain Aware Neural Dialog System
Figure 2 for Domain Aware Neural Dialog System
Figure 3 for Domain Aware Neural Dialog System
Figure 4 for Domain Aware Neural Dialog System

  Access Paper or Ask Questions

Enterprise to Computer: Star Trek chatbot


Aug 02, 2017
Grishma Jena, Mansi Vashisht, Abheek Basu, Lyle Ungar, João Sedoc

Figure 1 for Enterprise to Computer: Star Trek chatbot
Figure 2 for Enterprise to Computer: Star Trek chatbot
Figure 3 for Enterprise to Computer: Star Trek chatbot
Figure 4 for Enterprise to Computer: Star Trek chatbot

  Access Paper or Ask Questions

Deriving Verb Predicates By Clustering Verbs with Arguments


Aug 01, 2017
Joao Sedoc, Derry Wijaya, Masoud Rouhizadeh, Andy Schwartz, Lyle Ungar

Figure 1 for Deriving Verb Predicates By Clustering Verbs with Arguments
Figure 2 for Deriving Verb Predicates By Clustering Verbs with Arguments
Figure 3 for Deriving Verb Predicates By Clustering Verbs with Arguments
Figure 4 for Deriving Verb Predicates By Clustering Verbs with Arguments

  Access Paper or Ask Questions

Latent Human Traits in the Language of Social Media: An Open-Vocabulary Approach


May 22, 2017
Vivek Kulkarni, Margaret L. Kern, David Stillwell, Michal Kosinski, Sandra Matz, Lyle Ungar, Steven Skiena, H. Andrew Schwartz

Figure 1 for Latent Human Traits in the Language of Social Media: An Open-Vocabulary Approach
Figure 2 for Latent Human Traits in the Language of Social Media: An Open-Vocabulary Approach
Figure 3 for Latent Human Traits in the Language of Social Media: An Open-Vocabulary Approach
Figure 4 for Latent Human Traits in the Language of Social Media: An Open-Vocabulary Approach
* In submission to PLOS One 
  Access Paper or Ask Questions

Semantic Word Clusters Using Signed Normalized Graph Cuts


Jan 20, 2016
João Sedoc, Jean Gallier, Lyle Ungar, Dean Foster

Figure 1 for Semantic Word Clusters Using Signed Normalized Graph Cuts
Figure 2 for Semantic Word Clusters Using Signed Normalized Graph Cuts
Figure 3 for Semantic Word Clusters Using Signed Normalized Graph Cuts
Figure 4 for Semantic Word Clusters Using Signed Normalized Graph Cuts

  Access Paper or Ask Questions

Two Step CCA: A new spectral method for estimating vector models of words


Jun 27, 2012
Paramveer Dhillon, Jordan Rodu, Dean Foster, Lyle Ungar

Figure 1 for Two Step CCA: A new spectral method for estimating vector models of words
Figure 2 for Two Step CCA: A new spectral method for estimating vector models of words
Figure 3 for Two Step CCA: A new spectral method for estimating vector models of words
Figure 4 for Two Step CCA: A new spectral method for estimating vector models of words
* Appears in Proceedings of the 29th International Conference on Machine Learning (ICML 2012) 
  Access Paper or Ask Questions

Transfer Learning Using Feature Selection


May 25, 2009
Paramveer S. Dhillon, Dean Foster, Lyle Ungar

Figure 1 for Transfer Learning Using Feature Selection
Figure 2 for Transfer Learning Using Feature Selection
Figure 3 for Transfer Learning Using Feature Selection
Figure 4 for Transfer Learning Using Feature Selection
* Masters' Thesis 
  Access Paper or Ask Questions