This repository contains the data and the code for reproducing the results of the paper: Listen and tell me who the user is talking to: Automatic detection of the interlocutor’s type during a conversation. Only the intermediate extracted features from raw data that are available for this repository as described in the paper, they can be downloaded from https://drive.google.com/drive/folders/1vFXsOdpKmfP-mBS-ZvMuKKTdYD59MCqx?usp=sharing. To get the raw data, i.e., conversational multimodal signals of 25 subjects, please contact us.
- Using linguistic features:
python src/train_models.py -t ling -m att-lstm- Using filter banks features
python src/train_models.py -t ling -m att-lstm- Remarque: use python src/train_models.py -h to train more models.
- Usinglinguistic features
python src/test_models.py -t ling -m att-lstm- Using filter banks features
python src/test_models.py -t ling -m att-lstm