DialogueCRN
Source code for ACL-IJCNLP 2021 paper "DialogueCRN: Contextual Reasoning Networks for Emotion Recognition in Conversations".
Quick Start
Requirements
python==3.6.10
torch==1.4.0
torch-geometric==2.0.1
torch-scatter==2.0.4
sklearn==0.0
numpy==1.19.5
pandas==0.24.2
Install related dependencies:
pip install -r requirements.txtDataset
The original datasets can be found at IEMOCAP, SEMAINE and MELD.
In this work, we focus on emotion recognition in textual conversations. Following previous works (bc-LSTM, DialogueRNN, DialogueGCN, et al.), raw features of textual modality are extracted by using TextCNN.
Training/Testing
For training model on IEMOCAP dataset , you can refer to the following:
EXP_NO="dialoguecrn_v1"
DATASET="iemocap"
WORK_DIR="${WORK_PATH}/DialogueCRN" # your work path
DATA_DIR="${WORK_DIR}/data/${DATASET}/IEMOCAP_features.pkl"
OUT_DIR="${WORK_DIR}/outputs/${DATASET}/${EXP_NO}"
python -u ${WORK_DIR}/code/run_train_ie.py \
--feature_type text --data_dir ${DATA_DIR} --output_dir ${OUT_DIR} \
--gamma 0 --step_s 3 --step_p 4 --lr 0.0001 --l2 0.0002 --dropout 0.2 --base_layer 2For training model on MELD dataset , you can refer to the following:
EXP_NO="dialoguecrn_v1"
DATASET="meld"
WORK_DIR="${WORK_PATH}/DialogueCRN" # # your work path
DATA_DIR="${WORK_DIR}/data/${DATASET}/MELD_features_raw.pkl"
OUT_DIR="${WORK_DIR}/outputs/${DATASET}/${EXP_NO}"
python -u ${WORK_DIR}/code/run_train_me.py \
--feature_type text --data_dir ${DATA_DIR} --output_dir ${OUT_DIR} \
--gamma 1.0 --step_s 3 --step_p 0 --lr 0.001 --l2 0.0002 --dropout 0.2 --base_layer 1
Run examples
bash ./script/run_train_ie.sh
bash ./script/run_train_md.shResult
Reproduced experiment results on th IEMOCAP and MELD datasets:
| Model | IEMOCAP | MELD | ||||
|---|---|---|---|---|---|---|
| Acc | w-F1 | ma-F1 | Acc | w-F1 | ma-F1 | |
| TextCNN | 49.35 | 49.21 | 48.13 | 59.69 | 56.83 | 33.80 |
| bc-LSTM+Att | 56.32 | 56.19 | 54.84 | 57.50 | 55.90 | 34.84 |
| DialogueRNN | 63.03 | 62.50 | 60.66 | 59.54 | 56.39 | 32.93 |
| DialogueGCN | 64.02 | 63.65 | 63.42 | 59.46 | 56.77 | 34.05 |
| DialogueCRN | 66.73 | 66.66 | 67.25 | 61.26 | 58.48 | 35.69 |
Citation
@inproceedings{DBLP:conf/acl/HuWH20,
author = {Dou Hu and
Lingwei Wei and
Xiaoyong Huai},
title = {DialogueCRN: Contextual Reasoning Networks for Emotion Recognition
in Conversations},
booktitle = {{ACL/IJCNLP} {(1)}},
pages = {7042--7052},
publisher = {Association for Computational Linguistics},
year = {2021}
}