Mood Detection

Proposal

Proposed dataset:
- DailyDialog Dataset: emtion distribution diverse
  - The types of emtions:
    - anger 5.87%
    - disgust 2.03%
    - fear 1%
    - joy 74.02%
    - sadness 6.61%
    - suprise 10.47%
    - Links
    - total of 13,118 samples
- SemEval Sentiment Analysis in Twitter
  - emtions tags are as follows:
    - anger: 23.97%
    - fear: 31.73%
    - joy: 22.7%
    - sadness: 21.6%
    - links
- emtion casue:
  - emtion tags are as follows:
    - happiness
    - sadness
    - suprise
    - disgust
    - links
Proposed architecture
- Bert + classification module
  - testable for classification variation:
    - resnet
    - fully connected layers