Check Point Report: https://reurl.cc/1Z7ADp
Video: https://reurl.cc/e372xM
Chieh-Yu Chuang, Chun-Pei Chen
National YangMing ChiaoTung University Hsinchu, Taiwan
The topic idea comes from our enthusiasm towards color science and psychological tests. We found out that the psychology tests nowadays restrict their participants to answer particular questions to get an customized output. However, since people's emotions and personality are quite complicated, we think it would be better to accept a wider range of inputs to analyze the results better. Also, we believe that dialogues may not contain only one type of emotions, it could be complicated combinations. Therefore, we chose to show our results via colors.
Environment: Google Colaboratory
Packages version: refer to requirements.txt in our github
Dataset: https://reurl.cc/VDQxYn
Happy 35%
Sadness 31%
Anger 15%
Fear 14%
Surprise 5%
Epoch/iteration: 10
Learning Rate: 0.01
TrainData, ValidationData, and TestData proportion: 6-1-3
Multinomial Logistic Regression (MLR)
Gate Recurrent Unit (GRU)
Long Short-Term Memory (LSTM)
Bidirectional Long Short-Term Memory (Bi-LSTM)
Chieh-Yu Chuang: 50%, Chun-Pei Chen: 50%