Torsten Wörtwein

About Me

I’m currently working at ETS on two main projects:

1. creating and deploying models to infer workplace skills of users and requirements of jobs; and
2. developing a smartwatch app that provides real-time speech feedback to the wearer.

During my Ph.D. in Language and Information Technology at CMU, I worked with Louis-Philippe Morency on multimodal machine learning applied to human behaviors.

Interests

My expertise is application-inspired machine learning with the goal of socially aware AI technologies that benefit people by combining information from multiple modalities (through computer vision, speech processing, and natural language processing). One of my main interests is making AI technologies more transparent so that users can understand them better.

Examples of socially aware technologies that I worked on include:

1. mental health: developing behavior markers for symptoms of mental health;
2. public speaking: automated estimation of social skills to improve public speaking; and
3. affective states: automated detection of sentiment, emotions, and mood.

Selected Peer-Reviewed Publications

• SMURF: Statistical Modality Uniqueness and Redundancy Factorization

  Torsten Wörtwein, Nicholas B. Allen, Jeffrey F. Cohn, and Louis-Philippe Morency
  ACM International Conference on Multimodal Interaction. ICMI 2024
  [Paper] [Code]

• Towards Improving Transparency in Multimodal Affect Perception

  Torsten Wörtwein
  Committee: Louis-Philippe Morency, Jeffrey F. Cohn, Florian Metze, and Emily Mower Provost
  [Dissertation]

• Neural Mixed Effects for Nonlinear Personalized Predictions

  Torsten Wörtwein, Nicholas B. Allen, Lisa B. Sheeber, Randy P. Auerbach, Jeffrey F. Cohn, and Louis-Philippe Morency
  ACM International Conference on Multimodal Interaction. ICMI 2023
  [Paper] [Code]

• Beyond Additive Fusion: Learning Non-Additive Multimodal Interactions

  Torsten Wörtwein, Lisa B. Sheeber, Nicholas B. Allen, Jeffrey F. Cohn, and Louis-Philippe Morency
  Findings of the Association for Computational Linguistics: EMNLP 2022. F-EMNLP 2022
  [Paper] [Code]

• Human-Guided Modality Informativeness for Affective States

  Torsten Wörtwein, Lisa B. Sheeber, Nicholas B. Allen, Jeffrey F. Cohn, and Louis-Philippe Morency
  ACM International Conference on Multimodal Interaction. ICMI 2021
  [Paper] [Code]

• Simple and Effective Approaches for Uncertainty Prediction in Facial Action Unit Intensity Regression

  Torsten Wörtwein and Louis-Philippe Morency
  IEEE International Conference on Automatic Face and Gesture Recognition. FG 2020
  [Paper] [Code]

• Computational Analysis of Acoustic Descriptors in Psychotic Patients

  Torsten Wörtwein, Tadas Baltrusaitis, Eugene Laksana, Luciana Pennant, Elizabeth S. Liebson, Dost Öngür, Justin T. Baker, and Louis-Philippe Morency
  Conference of the International Speech Communication Association. INTERSPEECH 2017
  [Paper]

• What Really Matters - An Information Gain Analysis of Questions and Reactions in Automated PTSD Screenings

  Torsten Wörtwein and Stefan Scherer
  IEEE Conference on Affective Computing and Intelligent Interaction. ACII 2017
  [Paper]

• Mobile Interactive Image Sonification for the Blind

  Torsten Wörtwein, Boris Schauerte, Karin Müller, and Rainer Stiefelhagen
  International Conference on Computers Helping People with Special Needs. ICCHP 2017
  [Paper]

• Multimodal Public Speaking Performance Assessment

  Torsten Wörtwein, Mathieu Chollet, Boris Schauerte, Louis-Philippe Morency, Rainer Stiefelhagen, and Stefan Scherer
  ACM International Conference on Multimodal Interaction. ICMI 2015
  [Paper]

• Automatic Assessment and Analysis of Public Speaking Anxiety: A Virtual Audience Case Study

  Torsten Wörtwein, Louis-Philippe Morency, and Stefan Scherer
  IEEE Conference on Affective Computing and Intelligent Interaction. ACII 2015
  [Paper]

• Exploring Feedback Strategies to Improve Public Speaking: An Interactive Virtual Audience Framework

  Mathieu Chollet*, Torsten Wörtwein*, Louis-Philippe Morency, Ari Shapiro, and Stefan Scherer
  ACM International Joint Conference on Pervasive and Ubiquitous Computing. UbiComp 2015
  [Paper]

• Color Decorrelation Helps Visual Saliency Detection

  Boris Schauerte, Torsten Wörtwein, and Rainer Stiefelhagen
  IEEE International Conference on Image Processing. ICIP 2015
  [Paper]