APSIPA-SER-with-A-and-T

This code is the implementation of Speech Emotion Recognition (SER) with acoustic and linguistic features. The network model is Convolutional Neural Network (CNN) + Bidirectional Long Short Term Memory (BLSTM) + Self-Attention and BERT. Before running this code, you should get model parameters from "APSIPA-SER-with-A" and "APSIPA-SER-with-T."

How to use

0. Run main.py in "APSIPA-SER-with-A" and "APSIPA-SER-with-T"
1. Edit hyper_param.yaml
2. Run main.py

python3 main.py

Paper

Ryotaro Nagase, Takahiro Fukumori and Yoichi Yamashita: ``Speech Emotion Recognition with Fusion of Acoustic- and Linguistic-Feature-Based Decisions, '' Proc. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pp. 725 -- 730, 2021.