Example Selected relevant parts of videos Transcribed videos LIWC analysis

Example selected relevant parts of videos transcribed

This preview shows page 31 - 34 out of 34 pages.

Example Selected relevant parts of videos Transcribed videos LIWC analysis provides word count Track non-verbal behaviors FACET: emotions, AUs OpenFace: head, gaze Extract data Pre- process data Data cleaning Data imputation Data integration
Image of page 31
Example Feature generation: formulating new features based on initial given features Initial features: happy, angry, sad, surprised, fear Generated feature: “Expressivity” Feature construction: creating new features by combining other features Control smile if AU 12 + any “control” muscle movements (e.g., dimples -AU14-, lip corner depressing -AU15-, chin raising –AU17-, lip tightening –AU23- and lip pressing -AU 24) Joyous smile if AU 12 + any “authentic” muscle movements (e.g., cheek raising –AU6-, mouth openness –AU25- and jaw dropping -AU26) Feature selection: decision on what subset of the initial given features should be used LASSO in the training set for automatic feature selection 32 Pre- process data
Image of page 32
Example Classifiers Liar or not Perceived by opponent as dishonest or honest Facial features, especially smile, predicted perception Multimodal approach is better predictor of actual lying Word count is used accurately International Conference on Multimodal Interaction 33 Analyze data Present results Publish data
Image of page 33
Today’s Topics 1. Data lifecycle overview 2. Steps in the data lifecycle 3. Example Automatically tracked dyadic behavior 34
Image of page 34

You've reached the end of your free preview.

Want to read all 34 pages?

  • Fall '17
  • Data Management, Computational Thinking and Data Science, Compression, Utility

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

Stuck? We have tutors online 24/7 who can help you get unstuck.
A+ icon
Ask Expert Tutors You can ask You can ask You can ask (will expire )
Answers in as fast as 15 minutes