multimodal language analysis in the wild cmu mosei
play

Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and - PowerPoint PPT Presentation

Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph Presenter: Paul Pu Liang Amir Zadeh, Paul Pu Liang, Jonathan Vanbriessen, Soujanya Poria, Edmund Tong, Erik Cambria, Minghai Chen, Louis-Philippe


  1. Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph Presenter: Paul Pu Liang Amir Zadeh, Paul Pu Liang, Jonathan Vanbriessen, Soujanya Poria, Edmund Tong, Erik Cambria, Minghai Chen, Louis-Philippe Morency 1 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  2. Progress of Artificial Intelligence Intelligent Robots and Multimedia Content Personal Assistants Virtual Agents 2 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  3. Continuous Theories of (Multimodal) Language Throughout evolution language and nonverbal behaviors developed together. Cries and Imitations Modern Language Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  4. Multimodal Language Modalities Language Visual Ø Lexicon Ø Gestures Ø Syntax Ø Body language Ø Pragmatics Ø Eye contact Ø Facial expressions Acoustic Ø Prosody Ø Vocal expressions 4 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  5. Multimodal Language Modalities Language Visual Sentiment Ø Positive Ø Lexicon Ø Gestures Ø Negative Ø Syntax Ø Body language Emotion Ø Anger Ø Pragmatics Ø Eye contact Ø Disgust Ø Facial expressions Ø Fear Ø Happiness Acoustic Ø Sadness Ø Surprise Ø Prosody Personality Ø Vocal expressions Ø Confidence Ø Persuasion Ø Passion 5 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  6. Multimodal Language Modalities Sentiment Language Visual Acoustic Emotion Personality Datasets Models 6 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  7. Multimodal Language Modalities Sentiment Language Visual Acoustic Emotion Personality Datasets Models 7 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  8. Multimodal Language Modalities Sentiment Language Visual Acoustic Emotion Personality Datasets Models ü Large-scale ü Diverse 8 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  9. Multimodal Language Modalities Sentiment Language Visual Acoustic Emotion Personality Datasets Models § Word-level alignment § Attention models § Memory-based models ü Large-scale ü Diverse 9 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  10. Multimodal Language Modalities Sentiment Language Visual Acoustic Emotion Personality Datasets Models § Word-level alignment § Attention models § Memory-based models ü Large-scale ü Good Performance ü Diverse ü Interpretable 10 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  11. Datasets for Multimodal Language § Require large and diverse amounts of data: § Diversity in samples Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  12. Datasets for Multimodal Language § Require large and diverse amounts of data: § Diversity in samples § Diversity in topics Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  13. Datasets for Multimodal Language § Require large and diverse amounts of data: § Diversity in samples § Diversity in topics § Diversity in speakers Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  14. Datasets for Multimodal Language § Require large and diverse amounts of data: § Diversity in samples § Diversity in topics § Diversity in speakers § Diversity in annotations Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  15. New Dataset: CMU-MOSEI 23,000 video segments 3 modalities 15 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  16. CMU-MOSEI Dataset 1,000 speakers 250 topics 16 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  17. Annotation Distributions 17 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  18. Annotation Distributions 18 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  19. Feature Extraction Language Sentiment Ø Positive Ø Glove word embeddings Ø Negative Visual Emotion Alignment Ø Anger Ø Facet features Ø Disgust Ø Word level Ø MultiComp OpenFace Ø Fear Ø P2FA Ø Face embeddings Ø Happiness Acoustic Ø Sadness Ø Surprise Ø COVAREP features MFCCs • Pitch tracking • 19 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  20. CMU-MOSEI Dataset Multimodal Language Audio-visual 20 Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  21. Models for Multimodal Language ! ! ! ! multimodal Multimodal Fusion Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  22. Models for Multimodal Language Interpretation ! multimodal § Importance of each modality § Interactions between modalities Multimodal Fusion

  23. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § Interactions between modalities unimodal " " ! ! # # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  24. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § Interactions between modalities bimodal unimodal " " ! ! # # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  25. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § Interactions between modalities trimodal bimodal unimodal " " ! ! # # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  26. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § ⊕ Interactions between modalities fusion weights trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  27. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § ⊕ Interactions between modalities fusion weights trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  28. Dynamic Fusion Graph (DFG) t = 1 $ multimodal ⊕ trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  29. Dynamic Fusion Graph (DFG) t = 1 t = 2 $ multimodal ⊕ trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  30. Dynamic Fusion Graph (DFG) t = 1 t = 2 t = 3 $ multimodal ⊕ trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  31. Dynamic Fusion Graph (DFG) t = 1 t = 2 t = 3 t = 4 $ multimodal ⊕ trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  32. Dynamic Fusion Graph (DFG) Interpretation Interpretation $ multimodal § Importance of each modality § Interactions between modalities fusion weights trimodal bimodal unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  33. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § Interactions between modalities fusion weights trimodal § Construction of bimodal and trimodal representations bimodal construction weights unimodal " ! # Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

  34. Dynamic Fusion Graph (DFG) Interpretation $ multimodal § Importance of each modality § Interactions between modalities fusion weights trimodal § Construction of bimodal and trimodal representations bimodal construction weights ! ",$ ! %,$ ! ",% unimodal ' & ( Paul Pu Liang Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph

Recommend


More recommend