This is an implementation of Dialogical Emotion Decoder presented in this year ICASSP 2020. In this repo, we use IEMOCAP as an example to evaluate the effectiveness of DED.
- The performance is better than shown in the paper because I found a little bug in the rescoring part.
- To test on your own emotion classifier, replace
data/outputs.pkl
with your own outputs.- Dict, {utt_id: logit} where utt_id is the utterance name in IEMOCAP.
pip3 install virtualenv
virtualenv --python=python3 venv
source venv/bin/activate
pip3 install -r requirements.txt
Currently this repo only supports IEMOCAP.
The definitions of the args are described in ded/arguments.py
. You can modify all args there.
python3 main.py --verbosity 1 --result_file RESULT_FILE
Results of DED with beam size = 5.
Model | UAR | ACC |
---|---|---|
Pretrained Classifier | 0.671 | 0.653 |
DED | 0.710 | 0.695 |